Report No. SR99-10-02
Determination of Emissions
Credit and Average Test Times
for IM147 Testing
prepared for:
U.S. Environmental Protection Agency
October 11, 1999
prepared by:
Sierra Research, Inc.
1801 J Street
Sacramento, California 95814
(916)444-6666
-------
Determination of Emissions Credit and
Average Test Times for IM147 Testing
prepared for:
U. S. Environmental Protection Agency
October 11, 1999
prepared by:
Philip L. Heirigs
Richard W. Joy
Michael J St. Denis
John M. Lee
Sierra Research, Inc.
1801 J Street
Sacramento, CA 95814
(916) 444-6666
-------
Disclaimer
Although the information described in this report has been funded wholly or in part by
the United States Environmental Protection Agency under Contract No. 68-C7-0051, it
has not been subjected to the Agency's peer and administrative review and is being
released for information purposes only. It therefore may not necessarily reflect the views
of the Agency and no official endorsement should be inferred.
-------
Determination of Emissions Credit and
Average Test Times for IM147 Testing
Table of Contents
page
SUMMARY 1
Background ; 1
Scope of Work 2
Test Data Used in This Study 3
Development of IM147 Cutpoints 4
Methods to Reduce Test Time 6
Disappearing Vehicles 9
INTRODUCTION 12
Background 12
Scope of Work 15
Organization of the Report 16
IM147 TEST DATA 17
Summary Statistics ! 17
Comparison of Failure Rates 20
IM147 CUTPOINT ANALYSIS 31
Cutpoint Selection 32
Modal Predictive Fast Pass Analysis 43
Modal Predictive Retest Analysis 46
Model Year Exemptions 53
Model Year Exemptions Over All Model Years 56
Summary . 57
DISAPPEARING VEHICLES 61
Definition of Data Groups 61
Comparison of I/M Data To RSD Observations .62
Comparison of Current Disappearing Vehicle Estimates to Previous Estimates .... 64
Conclusions 65
REFERENCES 67
-------
List of Tables
page
1-1 Evaluation of Retest Criteria with 101 Vehicle Sample Based
on Final Cutpoints and Exempting 1992+ Vehicles 8
1-2 Excess Emissions Identified by Model Years Exempted 9
1 -3 Average Test Time (seconds) By Standards and Test Time
Reduction Methodology for the EM 147 9
1-4 Test Time and Percent Excess Emissions Identified by the IM147
Test By Standards and Test Time Reduction Methodology 10
1-5 Fraction of Vehicles Observed in the RSD Database that
Initially Failed an I/M Test in Maricopa County 11
3-1 Summary of Triplicate Phase 2 Scores by Test Sequence 19
3-2 DEQ "Alternative #2" Full-Cycle IM240 Cutpoints 21
3-3 DEQ "Alternative #2" IM147 Cutpoints (g/mi) 22
3-4 Phase 2-Only and Full IM240 Failure Rates Based on
DEQ "Alternative #2" Cutpoints 23
3-5 Summary of Vehicles in the Triplicate Phase 2-Only Test Program
with Non-Matching Pass/Fail Results Based on ADEQ Alternative
#2 Cutpoints 25
4-1 Regression Coefficients of IM240 Emission Rates Versus
Phase 2 of the IM240 .32
4-2 Startup IM240 Cutpoints and Iml47 Cutpoints Developed in This Study
(Composite/Phase 2 Cutpoints in g/mi, IM240-IM147) 34
4-3 Intermediate IM240 Cutpoints and IM147 Cutpoints Developed in
This Study (Composite/Phase 2 Cutpoints in g/mi, IM240-IM147) 35
4-4 Final IM240 Cutpoints and IM147 Cutpoints Developed in This
Study (Composite/Phase 2 Cutpoints in g/mi, IM240-IM147) 36
4-5 Evaluation of Emissions Standards for the IM147 Impact on
Test Time and Excess Emissions Lost 37
4-6 Tests with Type I and II Errors (IM240 & Third IM 147 Test Results Differ ... 39
-ii-
-------
4-7 Comparison of Fast-Pass Effectiveness for the IM147 Impact
on Test Time and Excess Emissions Lost 45
4-8 Evaluation of Retest Criteria (Based on Final Cutpoints) 50
4-9 Comparison of Retest Predictive Algorithm Effectiveness for
the IM147 Impact on Test Time and Excess Emissions Lost 52
4-10 Evaluation of Retest Criteria with 101 Vehicle Sample Based
on Final Cutpoints and Exempting 1992+ Vehicles 53
4-11 Comparison of Exempting 1992+ Model Years (Current + 5)
Impact on Test Time and Excess Emissions Lost 55
4-12 Excess Emissions Identified versus Model Years Exempted 56
4-13 Effect of LDV Model Year Exemptions on Excess
Emissions Identification 59
4-14 Average IM147 Test Time (in seconds) 60
4-15 Percent Excess Emissions Identified by the IM147 Test 60
5-1 Fraction of Vehicles Observed in the RSD Database that
Initially Failed an I/M Test in Maricopa County 62
5-2 Fail-Fail Vehicles in the RSD Database that Continue to
Attempt to Pass an I/M Test 64
-iii-
-------
List of Figures
page
1-1 Test Sequence Used to Investigate Triplicate IM147
Tests in Arizona Test Lanes 4
3-1 Test Sequence Used to Investigate Triplicate IM147
Tests in Arizona Test Lanes 18
3-2 Model Year Distribution of Sample Fleet and AZ Overall Fleet 18
3-3 IM147 Trace Showing Phases 1 & 2 23
3-4 Second-by-Second CO Emissions from Triplicate Phase 2 Testing
1998 Pontiac Bonneville (Record No. 14) 26
3-5 Second-by-SecOnd CO Emissions from Triplicate Phase 2 Testing
1989 Dodge Dynasty (Record No. 15) 27
3-6 Second-by-Second CO Emissions from Triplicate Phase 2 Testing
1995 Toyota 4Runner (Record No. 24) 28
3-7 Second-by-Second CO Emissions from Triplicate Phase 2 Testing
1993 Ford Ranger (Record No. 23) 29
4-1 Failure Rates for Three Consecutive IM147 Tests Followed By an
IM240 Test By Vehicle Type Using Final Cutpoints 41
4-2 Failure Rates for Three Consecutive IM147 Tests Followed By an
IM240 Test By Model Year Group Using Final Cutpoints . . . 42
4-3 IM147 Test Modes Used for Fast-Pass Cutpoints Development .; 44
4-4 Modes Used For Development Of Retest Algorithms 47
4-5 IM147 Retest Predictive Model Logic LDGV 48
4-6 IM147 Retest Predictive Model Logic LDGT 1&2 49
4-7 Cumulative Excess IM240 Emissions Identified by Model Year, Based on EPA's
Final Cutpoints 58
-IV-
-------
1. SUMMARY
Background
Under the Clean Air Act Amendments of 1990, metropolitan areas with the most serious
air quality problems are required to implement so-called "enhanced" I/M programs. Two
different test procedures for exhaust emissions testing in enhanced programs have been
approved by EPA: the "IM240" test, and the "Acceleration Simulation Mode" (ASM)
test. With either procedure, the efficiency of the testing process depends on how quickly
accurate decisions can be made as to whether a vehicle should pass or fail.
Efficiency is important to both the cost and the public acceptability of I/M programs. For
this reason, methods that maintain the accuracy of the tests, but can reduce test time are
important. This is especially true for programs that have been operating for a number of
years and the capacity of the test networks is being consumed. The reduction in testing
capacity is being caused by several reasons including (1) the increased failure rates that
will occur as more stringent cutpoints are implemented, causing fewer fast-passes (i.e.,
longer tests) and also more after repair retests; and (2) increases in the number of vehicles
subject to testing (the fleet is growing). Unless measures are taken in current programs to
offset the overall impact of these factors, a substantial increase in the existing network
capacities will be needed to keep wait times at acceptable levels. This, of course, would
translate into increased per-test fees. The main goal of the present work was to evaluate
means of reducing test times while maintaining the emissions benefits associated with the
IM240 test.
The most significant challenge to reducing test time is the potential for false failures as a
result of inadequate preconditioning. Before the vehicle is thoroughly warmed up, high
emissions can be caused by air-fuel ratio enrichment or an inactive catalytic converter. In
addition, increased emissions due to purging of loaded canisters may also be an issue
associated with inadequate preconditioning prior to I/M testing. In back-to-back IM240
testing of 336 vehicles conducted in 1997 by Gordon-Darby at its test lanes in Arizona,
19% of the 1981 and later model passenger cars and light trucks failing the initial IM240
test (based on the startup cutpoints) passed when immediately retested; 6% would have
passed the final cutpoints on the retest after failing the startup cutpoints on the initial
test.1* For these reasons, attempts to shorten the test length need to carefully evaluate the
effects of inadequate preconditioning on false failures.
During 1996 and 1997, Sierra conducted evaluations of preconditioning requirements
using data obtained from samples of vehicles recruited from IM240 lanes in Phoenix,
Arizona, and a laboratory test program at Sierra's facilities in Sacramento, California.
*
Superscripts denote references listed in Section 6 of this report.
-1-
-------
The conclusions from the two evaluations were that (1) a vehicle needing further
preconditioning can be identified through modal analysis of the emissions recorded
during the test; and (2) using the more aggressive portion of the IM240 (the last 147
seconds, often referred to as "Phase 2" ) may warm vehicles up faster while reducing test
time.
Sierra performed an initial analysis of the IM240 Phase 2-only test option under contract
to the State of Arizona Department of Environmental Quality (DEQ).2 That study was
based on a combination of data from the 2% random test sample (consisting entirely of
full-duration tests) routinely collected in the Arizona IM240 program and a limited
number (101 tests) of triplicate (back-to-back-to-back) tests of Phase 2 of the IM240
(herein after called the "IM147") conducted as part of the 1997 EPA evaluation.
Cutpoints were developed for the full IM147 and the second half of the IM147 (seconds
67 to 147) to complement the full IM147 cutpoints. Fast-pass cutpoints for both the
entire IM147 and Phase 2 of the IM147 were also developed. The present study was
developed as a follow-up to verify the credit level estimates of the previous work and also
provide a more robust estimate of average IM147 test times.
As the data from the test lanes were being analyzed to evaluate repeat or return tests, it
was discovered that a significant fraction of vehicles never passed their final IM240 test.
Even accounting for the issuance of program waivers, the incomplete repair rate ranged
from 5% to 27% depending on model year range and vehicle type. Several possibilities
were theorized to account for the high incomplete repair rate, including vehicles being
scrapped, sold out of the area, registered illegally out of the area but continuing to operate
within the area, or driven illegally without current registration. Part of the current study
was to determine the fate of these vehicles with excess emissions to determine if they are
still operating on the road.
Scope of Work
For this study, Sierra was to assist in the development and implementation of a testing
program to collect sufficient data to throughly evaluate the IM147 relative to the IM240.
Data were collected from over 300 randomly selected light-duty cars and trucks arriving
at the test lane during normal queuing conditions. Data collected include triplicate Phase
2 test results, followed immediately by a full-duration IM240 test.
The first step in the analysis of the testing data was to use the data collected under Task 1
to revise the fast-pass cutpoints and algorithms previously developed for IM147 testing.
Startup, alternative (or intermediate), and final standards were developed for both the full
IM147 test and Phase 2 (the second half of the test). Excess emissions were calculated
over the range of IM147 cutpoints developed by Sierra by summing the excess emissions
The second portion of the IM 240 (seconds 94 to 239) is herein after called the "end portion of the
IM240." If those 147 seconds are used alone as a single test as opposed to the second portion of the
IM240, it is designated as the "IM147." This has been done for clarification because "Phase 2" of the
IM147 is discussed later in the report, which represents seconds 67 to 147 of the IM 147 test. This will be
referred to as "Phase 2."
-2-
-------
identified for the test fleet by the IM147 test and dividing by the sum of the excess
emissions identified for the same vehicles by the subsequent IM240 test. The effect on
excess emissions identification due to exempting up to 10 model years was also
quantified. At the same time, the effect on average test times was estimated for the test
fleet for each of the IM147 conditions considered. To determine the fate of the
"disappearing" vehicles, Sierra analyzed historical I/M data combined with remote
sensing data to quantify the air quality benefit impact of failing vehicles not receiving a
passing IM240 test. The disposition of disappearing vehicles was evaluated by analyzing
remote sensing data to see if these vehicles were still operating in the enhanced area.
Based on the results of the above investigation, an analysis was conducted of the effect of
such illegal vehicle operation on excess emissions. Suggested solutions are provided to
address the problem from both a practical program implementation standpoint and a SIP
credit allocation perspective.
Test Data Used in This Study
Gordon-Darby conducted the test program at its I/M lanes in Phoenix, Arizona, from
March 2 to March 17, 1998. That program included 304 vehicles (193 cars and 111 light-
duty trucks) tested over triplicate IM147 tests followed by a full IM240 test, as shown in
Figure 1-1. The selection process used resulted in vehicles waiting in a queue for
approximately 5 to 15 minutes prior to testing.
Initial analysis of the data reveals substantial reductions in mean IM147 scores between
the first and second test in the sequence, with lesser reductions occurring between the
second and third test in the sequence. There were a few vehicles in the sample that had
inconsistent emissions across the series of four tests. There were 24 vehicles that fell into
this group but there were four vehicles that stood out because they all had significantly
higher HC and CO emissions during the full IM240 procedure than during the first three
IM147 tests. After detailed analysis of the modal (second-by-second) data from each of
these four vehicles, the cause of the inconsistent emissions from these vehicles could not
be satisfactorily explained. Because the only IM240 used for evaluation of the IM147 is
at the end of the testing, the simplistic analysis of excess emissions identified by the
IM147 test using the final IM240 as the "gold standard" will cause IM147 testing to look
poor by comparison since it would falsely pass several vehicles with very high CO
emissions. However, it is apparent that the higher emissions certain vehicles exhibited
during the IM240 are due to some anomaly and not to a fundamental problem with the
IM147 trace. If these vehicles had been tested over three IM240 tests, they most likely
would have also exhibited similar behavior. For these reasons, these four vehicles were
removed from the analysis because equivalent (non-preconditioned) IM240 emissions
information for these vehicles could not be determined.
-3-
-------
Figure 1-1
Test Sequence Used to Investigate Triplicate IM147 Tests in Arizona Test Lanes
5-15rrinutes First Second Third Hot
in queue IM147 test IM147test IM147 test IM240Test
Time
Development of IM147 Cutpoints
For the IM147, startup, intermediate, and final cutpoints were developed, as well as "two
ways to pass" cutpoints. For each set of standards, an assessment of the excess emission
losses using the IM147 versus the IM240 was performed and the potential effect on test
time from using the IM147 was evaluated. The test time modeling was performed with
the sample test fleet data using the IM147 (with a maximum of two retests) versus using
the IM240 (with a maximum of one retest).
Next, methods to reduce the test time were evaluated. As mentioned previously, if test
time can be reduced, this is equivalent to increasing the testing network's capacity at no
additional cost. In addition, reductions in test time can also help to reduce wait times for
consumers. Previous studies by Sierra have investigated several techniques that can be
applied to the test data from vehicles while they are being tested to reduce the overall test
time, with only minor losses in the excess emissions identified. The first method
developed was fast-pass standards to allow very clean vehicles to exit the test early. The
next method developed was retest algorithms that can predict if a failing vehicle would
pass if retested. If a vehicle would not pass if retested, then retesting the vehicle is only
increasing the test time with no benefit. The last technique evaluated was model year
exemptions. Exempting newer model year vehicles from testing for their first four or five
-4-
-------
years can significantly reduce total test time for'the fleet. Since new vehicles rarely fail
I/M tests, the excess emissions lost from not testing these vehicles should be small.
Each of these techniques to reduce test time was also evaluated for its impact on reducing
test time at the cost of potentially losing some excess emissions. The testing of each
technique built on the previous technique, i.e., testing of the retest algorithms was done
while first applying the fast pass criteria. There were several instances where this
building- block approach produced unexpected but beneficial results.
Cutpoint Selection - Development of the cutpoints for the IM147 was based on linear
regressions between full (composite) IM240 emission rates and emission rates for the
IM147 portion of those same IM240 tests by model year group. The IM240 test data
used were from the 304 vehicles described earlier with the four outliers removed from the
data set. Using the regressions and then modifying the resulting cutpoints for
discontinuities, startup, intermediate, and final cutpoints were developed for LDGVs,
LDGTls, and LDGT2s. The intermediate cutpoints were based on simply applying the
regression equations to the average of the IM240 composite startup and the final
cutpoints. The resulting standards are presented in Section 4 of this report.
To evaluate the impact of the new cutpoints on test time and excess emissions identified,
the modal data from the 300 vehicles used to develop the standards were "tested" using
the two-way-to-pass methodology for all three IM147s and the IM240 at the end of the
testing pattern. The IM240 was used as the standard for comparing excess emissions
identified, with excess emissions defined as emissions above the cutpoints on the IM240.
If a vehicle failed the IM240 and the vehicle also failed the IM147, then the IM147 would
be attributed with capturing the excess emissions from that vehicle, and those emissions
would have a value equal to the IM240 emissions over the IM240 standard. If the IM147
did not fail a vehicle that failed the IM240, the excess emissions from the IM240
(emissions over the IM240 standard for the IM240 test) were considered excess emissions
not identified. The ratio of excess emissions for vehicles failing the IM147 versus
vehicles failing the IM240 is the percent of excess emissions identified.
Analysis of the new cutpoints shows there were a few excess emissions lost with the use
of the IM147 versus the IM240. For startup standards, the IM147 identified all of the
excess emissions the IM240 identified and only lost 0.8% of the excess HC emissions
with the intermediate standards. The maximum loss was 1.8% of excess emissions for
CO using the final standards. On an individual pollutant basis, HC and CO failure rates
for LDTls were underpredicted, while for NOx the failure rates were slightly
overpredicted. LDT2s predicted well for CO from the first test; HC and NOx were
overpredicted for the first test, but agreed well with IM240 failure rates for the third
IM147 test. Looking at the failure rates by model year showed a similar trend between
the pollutants, in that the first IM147 test failure rate is about 30% to 40% higher than the
IM240, but agrees well after the second IM147 test. For 1981-1985 model year vehicles,
the agreement for HC is good by the second IM147 test; however, the EM 147
underpredicts the failure rates for CO, and the agreement for NOx is not good until the
third IM147 test. For 1986-1989 model year vehicles, the third IM147 test for HC
underpredicted failure rates and the second IM147 test overpredicted failure rates. NOx
failure rates were overpredicted even for the third test. The failing sample for 1990-1993
-5-
-------
model year vehicles was small, but the agreement was good after the third test for all
three pollutants, with the largest variation in NOx emissions for the first IM147 test.
There were too few failing 1994 and newer vehicles to determine if the cutpoints agree
well.
Review of the test times show a 20% increase in test time going from startup cutpoints to
final cutpoints. As expected, test times by model year group were higher for older model
year vehicles, since these vehicles would have higher emissions and be more likely to fail
the test, requiring another test.
Methods to Reduce Test Time
Fast-Pass - The first technique applied to the new IM147 test to help shorten test time
was to apply fast pass standards to allow those vehicles that are functioning well below
the standards to complete the test early. Development of fast-pass standards is based on
regression models of the mean emissions of passing vehicles plus two times the standard
error at 13 modes in the test. In all, 18 regression models were developed: six each for
HC, CO, and NOx, with these six representing composite and the end portion (seconds 67
to 147) of the IM147 (herein after referred to as "Phase 2") standards for LDGVs,
LDGTls, and LDGT2s. For the IM147, the last 13 modes developed for the IM240 in
the previous study were used and the last four of these modes were used for the Phase 2
portion.
Again, the vehicles from this study were used to calculate excess emissions identified and
change in test time both with and without fast-pass regression. This analysis was
performed using the startup, intermediate, and final IM240 HC, CO, and NOx standards.
The following overall reductions in test time were achieved: 58% for startup cutpoints,
51% for intermediate cutpoints, and 41% for final cutpoints. Comparison of these test
times to test times for the use of fast-pass for IM240 testing previously reported to EPA
shows the IM147 has lower test times for startup and intermediate cutpoints; however,
test times for final cutpoints are almost the same.
Overall, the IM147 identified over 95% of excess emissions. There was, however,
variation by model year group. The majority of the loss in excess emissions came from
1985 to 1989 model year LDGTs, the only group having an excess emissions
identification rate below 95%.
Modal Predictive Retest Analysis - The second method of reducing test time involved the
development of algorithms to predict if a vehicle would benefit from a retest due to lack
of proper preconditioning. If a vehicle would continue to fail repeated tests then retesting
the vehicle would be an inefficient use of testing time. If failing vehicles that will
continue to fail can be discriminated from those vehicles that would benefit from another
test (vehicles that failed due to a lack of preconditioning but would pass an additional
test), retesting of the fail/fail vehicles could be avoided and average test time could be
reduced.
-6-
-------
The composite emission rate, the Phase 2 emission rate, and the concentrations during
four sections of the test were evaluated to determine what relationships exist in the data
that can help predict if a vehicle would benefit from a retest. Due to limited sample size,
LDGTls and LDGT2s were combined for this analysis and failures between either IM147
test one and two, or IM147 test two and three, were treated the same for developing the
retest criteria. Using these conditions, data from 111 failed tests (75 LDGTs and 36
LDGTs) were used for development of the retest algorithms.
Criteria development was an empirical process, involving manual evaluation of the
relationships between all the data. Under the first criterion applied, a vehicle would not
be retested if it failed for all three pollutants. Additional criteria used the relationships
between the entire test or Phase 2 of the test relative to the emissions standards or the
ratio of concentrations between mode 1 and mode 4. Applying these criteria in
combination to all of the vehicles in the sample fleet that failed, the overall ability for the
algorithms to correctly predict a vehicle needing a retest was 90%; however, it was better
for LDGTs than for LDGVs.
Addition of the retest algorithms showed significant reductions in test time. The use of
fast-pass and retest algorithms showed overall reductions in test time of 68% (from 172
seconds per test to 55 seconds per test) for startup cutpoints, 64% (from 185 seconds per
test to 66 seconds per test) for intermediate cutpoints, and 61% (from 206 seconds per test
to 81 seconds per test) for final cutpoints. At the same time, the impact on excess
emissions lost was minor. There were several cases where the retest logic prevented a
vehicle from undergoing a second test and then being falsely fast-passed, which actually
prevented excess emissions from being lost.
To more objectively evaluate the accuracy of the retest predictive algorithms, they (along
with the fast-pass standards) were applied to another sample of 101 vehicles tested on the
IM147 in previous study.3 The testing evaluated the false-pass or false-fail occurrences,
excess emissions lost, and average test times. The results are shown in Table 1-1.
The results show that the number of errors for LDGTl&2s were larger than for LDGVs.
This is most likely due to the LDGT1&2 retest criteria being based on fewer data than the
LDGV retest criteria.
The percent of excess emissions identified shows the largest emissions losses for HC,
then CO. All NOx emissions were identified. The losses in excess HC emissions
identified are much higher than for HC in the 304-vehicle sample; however, at the same
time, the excess emissions NOx loss is lower and the CO loss is in the range of the other
measurements.
-7-
-------
Table 1-1
Evaluation of Retest Criteria with 101 Vehicle Sample
Based on Final Cutpoints and Exempting 1992+ Vehicles
Sample Size
(pre-1992 MY)
Percent
Correct
Retested
When
Vehicle
Would Still
Fail
Did Not
Retest
When
Vehicle
Would Pass
Percent of Excess
Emissions
Identified
HC
CO
NOx
LDGV
31
80.6 %
3.2 %
16.1 %
94.1
98.6
100
LDGT1&2
14
64.2 %
14.3 %
21.4%
30.6
30.4
71.1
Weighted average
of LDGV and
LDGT1&2
45
75.6 %
6.6 %
17.8%
74.2
89.8
92.1
Model Year Exemptions - A very efficient method to reduce the overall average test time
is to simply remove some vehicles from testing. Currently, EPA has published draft
guidance4 on the use of "Clean Screening" methods for identifying potentially clean
vehicles to be exempted from testing. The methods discussed in the report and currently
under evaluation include the use of model year exemptions, a low emitter profile, and
remote sensing to identify likely clean vehicles.
Of the proposed methods, model year exemptions are the easiest to implement and should
have virtually no direct cost to implement. Part of the reason for developing methods to
reduce test time is to improve the efficiency and cost effectiveness of emissions testing.
For this reason, model year exemptions seem to be a very reasonable method to use and
were considered as part of the present study.
The first comparison of the potential excess emissions lost from model year exemptions
was performed by removing the excess emissions that occurred in the IM240 from those
identified by the IM147, for the model years being considered for exemption. Vehicles in
the present study ranged from 1981 to 1998 model years. However, because only one
1998 model year vehicle was in the database, it was assumed that a "current plus five"
model year exemption would include exempting 1997 through 1992 model year vehicles.
Reductions in total test time using fast-pass cutpoints, retest algorithms and exempting
the current plus five model years reduced test times by 82% (from 172 seconds per test to
31 seconds per test) for startup cutpoints, 78% (from 185 seconds per test to 41 seconds
per test) for intermediate cutpoints, and 73% (from 206 seconds per test to 55 seconds per
test) for final cutpoints. For startup cutpoints, the excess emissions lost was greatest for
CO; for the intermediate cutpoints, NOx was highest. For final cutpoints, the maximum
excess emissions lost was for CO at 7.9%. Table 1-2 summarizes the excess emissions
identified by model years exempted, for the final IM147 cutpoints. For exempting 1994+
vehicles, the excess emissions lost is less than 2% for each of the three pollutants.
-8-
-------
Table 1-2
Excess Emissions Identified by Model Years Exempted
Based on Final IM247 Cutpoints
HC
CO
NOx
1994 +
99.6 %
98.2 %
99.9 %
1993 +
99.6 %
98.2 %
92.6 %
1992 +
96.2 %
92.6 %
92.6 %
1991 +
87.1 %
88.4 %
72.8%
Summary of Test Time Changes - Table 1-3 summarizes the changes in test time as each
of the methods to reduce test time was applied. As the table shows, up to a 73%
reduction in test time can be achieved from the final standards when using the test time
reduction techniques, if the exempted 1992+ model-year vehicles are included in the
analysis. When only those vehicles arriving at the test lanes are considered, the possible
reduction in test time drops to 44%. The largest reduction was for use of fast-pass, which
is probably due to the fact that many vehicles are very clean and can get out of the
emissions test quickly.
Table 1-3
Average Test Time (seconds) by Standards and Test Time
Reduction Methodology for the IM147
Startup Cutpoints
Intermediate Cutpoints
Final Cutpoints
Cutpoint only, two possible retests
172
185
206
Added fast-pass (
72
92
121
Added retest algorithm .
•55
66
81
Added exemption
of 1992+
model years
All Vehicles
31
41
55
Non-Exempt
Vehicles
65
86
115
Overall %
reduction
in test time
All Vehicles
82%
78%
73%
Non-Exempt
Vehicles
62%
54%
44%
Summary of Excess Emissions Identified - Excess emissions identified for each pollutant
are shown in Table 1-4 for each method to reduce test time and for each set of cutpoints
by pollutant. Overall, total excess emissions lost by switching to the IM147 are low, up
-9-
-------
Table 1-4
Test Time and Percent Excess Emissions Identified by the
IM147 Test By Standards and Test Time Reduction Methodology
Startup Cutpoints
Intermediate Cutpoints
Final Cutpoints
Test
Time
HC
CO
NOx
Test
Time
HC
CO
NOx
Test
Time
HC
CO
NOx
Cutpoint only, two
possible recesis
172
100
100
100
185
99.2
100
93.6
206
99.6
98.2
99.6
Added fast-pass
72
96.5
95.1
100
92
98.8
97.8
89.5
121
99.6
95.5
99.6
Added retest
algorithm
55
97.2
95.1
100
66
98.4
100
93.2
81
99.2
97.5
99.6
Added exempting
1992+ model years'
31
95.2
81.7
88.5
41
96.0
91.5
84.1
55
95.9
92.1
92.4
a Test time estimates are based on a weighted average for all vehicles, including a test time of 0 for
1992+ vehicles exempted from testing.
to the point where model years are exempted. However, for final outpoints, the IM147
retains over 92% of the excess emissions identified by the IM240 even when exempting
current plus five model years.
Disappearing Vehicles - To perform the analyses, data regarding vehicles subjected to
I/M testing in Maricopa County from July through September 1997 were used to identify
vehicles that initially passed or failed the I/M test. The initial test failures were tracked
through the end of the year to determine which of these vehicles had still not passed an
emissions test (Fail-Fail vehicles) within a three- to six-month time period following their
initial test date (i.e., by December 31, 1997). To determine whether the Fail-Fail vehicles
were still being operated in the area, the license plate numbers of these vehicles were
compared to license plate data of vehicles identified by remote sensors in Maricopa
County from January 1 through March 31, 1998. The percentages of vehicles observed in
the RSD data in Fail-Fail and Fail-Pass categories are presented in Table 1-5 by model
year groupings, as well as the ratio of the frequency that vehicles are seen on the road.
It was found that a smaller fraction of older vehicles were observed in the RSD data for
both the Fail-Fail and the Fail-Pass categories. However, for all of the model year ¦
groups, the fraction of Fail-Fail vehicles observed by RSD is less than that of the Fail-
Pass vehicles. About 50% of the Fail-Fail vehicles in the pre-1974 model year group
have been removed from the road but only 9% of the 1992 and later model year Fail-Fail
vehicles do not continue to operate on the road in Maricopa County. Waivered vehicles
were not responsible for this difference, accounting for only 0.71% of the disappearing
vehicles. In addition, further analysis indicates that about 20% of the Fail-Fail vehicles
operating in the area continue to be tested in an attempt to receive a passing I/M score.
-10-
-------
Table 1-5
Fraction of Vehicles Observed in the RSD Database
that Initially Failed an I/M Test in Maricopa County
Model Year
Group
Fail-Fail
July-Sept
1997
Fail-Fail
Observed
by RSD
Fail-Pass
July-Sept
1997
Fail-Pass
Observed
by^RSD
Ratio of.
Fail-Fail to
Fail-Pass
Pre-1975
823
2.7%
2,657
5.3%
51%
1975-1980
2,512
4.4%
8,690
7.0%
63%
1981-1984
2,148
6.2%
4,935
8.8%
70%
1985-1987
2,279
7.8%
6,662 .
10.2%
77%
1988-1991
1,154
9.2%
5,432
13.0%
71%
1992 +
282
13.8%
2,645
15.2%
91%
As noted above, the issue of "disappearing" vehicles was first identified in a study
performed by Sierra for DEQ. In that study, which was based on an analysis of the 2%
Random Sample IM240 database, initial test results from January 1996 to December
1996 were merged with after-repair tests conducted from January 1996 through April
1997. Although a slightly different methodology was used in that analysis, similar results
were obtained.
Summary - The analyses above indicate that vehicles that continue to fail the I/M test
after a number of months (i.e., the Fail-Fail "disappearing" vehicles) are observed
operating on the road less frequently than their counterparts that received an initial I/M
test in the same time period. However, this is a function of vehicle age, with older
vehicles being less likely to continue to be operated (possibly scrapped, parked, or sold
outside the area) than newer vehicles. Using RSD data to infer operation frequency, it
appears that about half of the pre-1975 model year Fail-Fail vehicles do not remain in
operation six to nine months after their initial test, while nearly 90% of the 1992 and later
model year Fail-Fail vehicles remain on the road. Of those vehicles that do remain on the
road, approximately 20% continue to attempt to pass the I/M test.
The fraction of initial test failures not receiving complete repairs (i.e., the Fail-Fail
vehicles) estimated in this analysis agreed very well with the results of Sierra's previous
analysis prepared for DEQ. Both analyses showed that older vehicles are more likely to
not be repaired completely, with about 30% of the 1981 to 1984 model year initial test
failures not receiving a passing score on the last I/M test.
###
-11-
-------
2. INTRODUCTION
Background
Under the Clean Air Act Amendments of 1990, metropolitan areas with the most serious
air quality problems are required to implement so-called "enhanced" I/M programs. One
element of an enhanced program is a more effective test procedure than the simple idle
tests used in "basic" I/M programs. Two different test procedures for exhaust emissions
testing in enhanced programs have been approved by EPA: the "IM240" test, and the
"Acceleration Simulation Mode" (ASM) test. Both of these procedures have been shown
to be capable of separating vehicles with excessive exhaust emissions from other
vehicles; however, the accuracy of the test depends on whether tested vehicles have been
adequately preconditioned and whether the speed-time profile associated with each test
procedure is closely followed. With either procedure, the efficiency of the testing process
depends on how quickly accurate decisions can be made as to whether a vehicle should
pass or fail.
Efficiency is important to both the cost and the public acceptability of I/M programs. For
this reason, methods that maintain the accuracy of the tests while also reducing test time
are important. This is especially true for programs that have been operating for a number
of years and have reached the design capacity of the inspection network. This can be
caused by several reasons, including (1) increased failure rates that will occur as more
stringent cutpoints are implemented, causing fewer fast-passes (longer tests) and also
more after-repair retests; and (2) increases in the number of vehicles subject to IM240
testing (i.e., the fleet is growing). Unless measures are taken in current programs to
offset the overall impact of these factors, a substantial increase in existing network
capacities will be needed to keep wait times at acceptable levels. The main goal of the
present work was to evaluate means of reducing IM240 test times while maintaining the
emissions benefits associated with this transient test.
The most significant challenge to reducing test time is the potential for a vehicle to be not
properly preconditioned and thus falsely fail the emissions test. For example, Indiana has
adopted the first 93 seconds of the IM240 trace, which it calls the IM93. Unfortunately,
this test cycle consists of the low-speed portion of the IM240 test, and may not be
aggressive enough (require enough power to drive) to fully warm up the vehicle.
Inadequate preconditioning of vehicles prior to testing is a potential cause of inaccurate
or inconsistent test results because exhaust emission levels depend on how thoroughly a
vehicle has been warmed up. Before the vehicle is thoroughly warmed up, high
emissions can be caused by air-fuel ratio enrichment or an inactive catalytic converter. In
addition, increased emissions due to purging of loaded canisters may also be an issue
associated with inadequate preconditioning prior to I/M testing. In back-to-back IM240
testing of 336 Arizona vehicles conducted in 1997 by Gordon-Darby, 19% of the 1981
-12-
-------
and later model passenger cars and light trucks failing the initial IM240 test (based on the
startup cutpoints) passed when immediately retested; 6% would have passed the final
cutpoints on the retest after failing the startup cutpoints on the initial test.5 For these
reasons, attempts to shorten the test length, such as the IM93, need to carefully evaluate
the effects of inadequate preconditioning, and address them in the analysis of the testing
results.
Under current EPA guidance, IM240 preconditioning procedures are woven into the
"two-ways-to-pass" standards. Vehicles that exceed the emissions standards established
for the entire 240-second test are passed or failed based on emissions occurring during the
last 147 seconds of the test. The separate set of standards that apply to the IM147 is
slightly more stringent. For vehicles that initially demonstrate high emissions, the first
93 seconds (Phase 1) of the test are used to precondition the vehicle for the second phase
of the test. In addition, EPA calls for a "second-chance" test whenever a vehicle fails the
initial test by less than 50% of the standard and was in a queue for more than 20 minutes
before being tested.
Considerable data have already been collected regarding the preconditioning
requirements for IM240 testing. During 1996 and 1997, Sierra conducted evaluations of
this issue using data obtained from samples of vehicles recruited from IM240 lanes in
Phoenix, Arizona, and a laboratory test program at Sierra's facilities in Sacramento. The
results of the 1996 analysis were reported in SAE Paper No. 962091. The 1997
evaluation also included an analysis of the effect on test duration of adopting EPA-
recommended "final" IM240 cutpoints. Preliminary conclusions from the two
evaluations are summarized below.
1. Using the current IM240 test procedures, it is estimated that 25% of the vehicles
failing the final IM240 standards would pass with further preconditioning.
2. Vehicles that would benefit from further preconditioning can be identified through
modal analysis of the emissions recorded during the IM240 test.
3. Two possible approaches to modifying the current preconditioning procedures
would be to:
a. retain existing IM240 test procedure and two-ways-to-pass standards, with the
entire IM240 to be repeated if the IM147 emissions failure is marginal,
emissions near end of the IM147 are relatively low, or emissions during the
IM147 are significantly lower than during Phase 1; or
b. eliminate Phase 1 and make the initial pass/fail decision based on running
only the IM147 (which is more aggressive and thereby possibly warms the
vehicles sooner), with a second-chance test (another IM147) for all vehicles
that initially fail, and a third-chance IM147 test if emissions during the
second-chance test are significantly lower than emissions during the initial
test.
-13-
-------
4. Adoption of final outpoints and more effective preconditioning procedures
involving a second full-IM240 (Option 3.a. above) will increase the portion of the
test involving dynamometer operation by more than 100%.
As a follow-on to the 1997 evaluation for EPA, Sierra subsequently conducted an
analysis for the Arizona DEQ of the effect on failure rates, I/M program benefits, and test
duration of the following changes to the current IM240 procedure: (1) implementation of
the Option 3.b. preconditioning procedures summarized above; (2) adoption of
"intermediate" cutpoints designed to maximize the CO emission reduction benefits being
achieved by the program; and (3) the exemption of either the first four or first five model
years from program requirements.
Due to concerns that past CO attainment demonstrations for Maricopa County have not
been realized due to inaccuracies in the MOBILE models used to generate emissions
projections, a non-MOBILE-based analysis methodology was used for the DEQ analysis.
To analyze the IM147 preconditioning option, Sierra used a combination of data from the
2% random test sample (consisting entirely of full-duration tests) that is routinely
collected in the Arizona IM240 program and a limited number (101 tests) of triplicate
(back-to-back-to-back) tests of the IM147 that were conducted as part of the 1997 EPA
evaluation. A key element of the analysis methodology involved the development of
cutpoints for the second half of the IM147 (Phase 2) to complement the full IM147
cutpoints. (The Phase 2 cutpoints were applied in a manner similar to the current IM240
procedure in which vehicles passing in Phase 2 are considered passing for the entire test.)
Fast-pass cutpoints for both the entire IM147 and Phase 2 were also developed.
A second issue of importance in the analysis was the discovery of a significant fraction of
vehicles failing on their initial test that never passed their final IM240 test. Even
accounting for the issuance of program waivers, the incomplete repair rate ranged from
5% to 27% depending on model year range and vehicle type. Discussions with DEQ staff
revealed that this incomplete repair rate had previously been identified as an issue of
concern based on an analysis conducted by Arizona's I/M contractor, Gordon-Darby.
According to DEQ staff, several possibilities have been theorized to account for the high
incomplete repair rate, including vehicles being scrapped, sold out of the area, registered
illegally out of the area but continuing to operate within the area, or driven illegally
without current registration. Notwithstanding this speculation, no data are currently
available to support any assumed distribution of the "disappearing" vehicles among these
potential outcomes. In the absence of such data, it was conservatively assumed that all
the disappearing vehicles (with the exception of those obtaining a program waiver)
continue to operate illegally within the program area. If an analysis can show that these
vehicles are not still operating on road, then there is less of a SIP credit loss from these
vehicles.
As noted above, while the 1997 EPA study involved the analysis of a considerable
amount of IM240 data, only a small subset (101 vehicles) was of use in projecting credit
levels and test times for an IM147 test program. An additional concern is that the 101-
vehicle study was not specifically designed to determine excess emission identification
rates and SIP credit levels. As a result, EPA only conditionally approved the emission
-14-
-------
credits developed by Sierra in the DEQ analysis. The present study was therefore
completed as a follow-up to verify this credit level and also provide a more robust
estimate of average IM147 test times.
Scope of Work
Work Assignment 0-02 called for Sierra to complete the following objectives:
1. Verify the preliminary excess emission identification rates and average test time
estimates that have been projected for the Arizona IM program from existing
IM147 and IM240 data sets collected in previous studies; and
2. Estimate the emission benefits lost due to failing vehicles not being retested per
the program specifications.
Four tasks were identified to accomplish these objectives.
Task 1, Test Plan Development and Data Collection - This task involved working with
Gordon-Darby to develop and implement a plan for collecting the test data needed to
complete Task 2, Excess Emissions Identification. Emissions data were collected from
over 300 randomly selected light-duty cars and trucks arriving at the test lane during
normal queuing conditions. Data collected include triplicate IM147 test results, followed
immediately by a full-duration IM240 test.
Data collection, driver participation incentives, and other program-related details were
performed under the guidance of DEQ and Gordon-Darby and were not Sierra's
responsibility. However, as part of this work assignment, Sierra reimbursed Gordon-
Darby for driver participation incentives and other costs associated with collecting the
required data. It was planned that the data would be submitted by Gordon-Darby to
Sierra on a regular basis to identify any potential problems (e.g., the lack of a
representative cross-section of vehicles subject to inspection); however, due to the speed
in which the data were collected, no feedback occurred.
Task 2. Excess Emissions Identification - After completion of the vehicle testing
described under Task 1, Sierra analyzed the resulting data. The analysis included an
evaluation of the effectiveness of the IM147 test procedures previously recommended by
Sierra relative to the current IM240 test procedure. The first step in the analysis was to
use the data collected under Task 1 to revise the fast-pass cutpoints and algorithms
previously developed for IM147 testing. Startup, alternative (or intermediate), and final
standards were developed for both the full IM147 test and Phase 2.
Using the data collected under Task 1, excess emissions were calculated over the range of
IM147 cutpoints developed by Sierra. The percent excess emissions identified were
calculated by summing the excess emissions identified for the test fleet by the IM147 test
and dividing by the sum of the excess emissions identified for the same vehicles by the
subsequent IM240 test. For the purposes of this calculation, IM240 excess emissions are
-15-
-------
defined as the difference between the vehicle's IM240 emissions and the respective EPA-
recommended final IM240 cutpoint for that make and model year. If this difference was
negative, i.e., the vehicle passes the IM240 cutpoint, the excess emissions are zero. The
IM147 excess emissions are defined in the same way; however, this calculation is done
only for those vehicles that fail the IM147 cutpoint. For each IM147 cutpoint
combination considered, the false failure and false pass rates were determined. The effect
of exempting up to 10 model years on excess emissions identification was also
quantified.
Task 3. Test Duration - Test times for IM147 testing were estimated for the test fleet.
The impact of using fast-pass IM147 cutpoints and the second-chance algorithms
developed based on the 1997 analysis for EPA and the subsequent DEQ analysis, which
were refined in this study under Task 2 above, were included in this evaluation. The
impact of exempting up to the first 10 model years on IM147 test times was also
quantified.
Task 4. Retest Non-Compliance - Sierra analyzed historical I/M data combined with
remote sensing data to quantify the air quality impact of failing vehicles never receiving a
passing IM240 score after being repaired. The disposition of such "disappearing"
vehicles was investigated by analyzing remote sensing data to determine the fraction of
those vehicles still operating in the enhanced area. Based on the results of the above
investigation, an analysis was conducted of the effect of such illegal vehicle operation on
excess emissions. Suggested solutions are provided to address the problem from both a
practical program implementation standpoint and from a SIP credit allocation
perspective.
Organization of the Report
Following this introduction, Section 3 describes data collection and the data set used in
the development of alternative cutpoints, failure rates, and test times for the IM147
testing. Section 4 describes the development of IM147 standards (composite and Phase 2
for startup, intermediate, and final use), the development of a modal predictive model for
determining fast pass, and the development of retest criteria for IM147 testing. The
excess emissions from the use of the final standards, the fast pass, and the retest criteria
are compared to the IM240 tests results. In addition, the effect on overall test time due to
the use of the new standards, fast pass model, and retest criteria was determined. The
addition of model year exemptions on excess emissions lost was determined for the first
ten model years. Section 5 presents the analysis of the fate of disappearing vehicles and
provides estimates of the excess emissions lost by disappearing vehicles not being
repaired. Section 6 lists the references cited in the report. A set of appendices provides
more detailed graphic and tabular presentations of data results.
###
-16-
-------
3. IM147 TEST DATA
Gordon-Darby conducted the test program under Task 1 of Work Assignment #0-02 at its
I/M lanes in Phoenix, Arizona from March 2 to March 17, 1998. The program included .
304 vehicles (193 cars and 111 light-duty trucks) tested over triplicate IM147 tests
followed by a full IM240 test. The test sequence is illustrated in Figure 3-1. The model
year distribution of the fleet is shown in Figure 3-2.
Each test center was given a $150 bonus for performing the testing, and all 10 different
I/M test centers located in the Phoenix area participated. One inspector and lane in each
test facility were designated as the study lane. Vehicles were tested when the facility
manager determined the alternative testing would not have a negative impact on wait
time. The criterion for this decision was that the number of vehicles per lane in the non-
study lanes needed to be less than four.
To ensure random selection of the test, the Study Inspector would scan the queue for the
closest white vehicle waiting in the lanes. If there were no white vehicles in the queue,
the Inspector would look for the palest, closest vehicle waiting in the lanes. The
inspector approached the first 1981 or newer vehicle following that vehicle, checked to
make sure the vehicle had at least one-half of a tank of gas, and asked the vehicle owner
if he or she was interested in participating in a study that would take approximately 30
minutes, for which they would receive $50. The selection process resulted in vehicles
waiting in a queue for approximately 5 to 15 minutes prior to testing. Most of the
vehicles participating in the program were receiving their initial test; however, 12
vehicles in the database were being re-tested after an initial failing score.
This section of the report summarizes the IM147 test results and discusses issues related
to the analysis of those results.
Summary Statistics
Table 3-1 presents a summary of the mean IM147 emission rates for the vehicles tested in
this program. Results are presented separately for cars versus light-duty trucks and for
the following model year groups:
• 1981 - 1985;
• 1986 - 1989;
• 1990- 1993; and
• 1994 and later.
-17-
-------
Figure 3-1
Test Sequence Used to Investigate Triplicate IM147 Tests in Arizona Test Lanes
in queue IM147 test IM147 test IM147 test IM240Test
Time
Figure 3-2
Model Year Distribution of Sample Fleet and AZ Overall Fleet
16%
14% -
12% -
10% -
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Model Year
-18-
-------
Table 3-1
Summary of Triplicate IM147 Scores by Test Sequence
Vehicle
MY
Sample
IM147
Phase 2 Test Score (g/mi)
Class
Group
Size
Sequence
HC
CO
NOx
LDV
81-85
20
1
1.27
14.85
2.84
2
0.94
13.20
2.46
3
0.86
13.50
2.35
IM240-Ph 2
0.83
12.47
2.28
86-89
40
1
0.72
8.87
1.54
2
0.47
7.26
1.33
3
0.44
8.34
1.24
IM240-Ph 2
0.49
10.70
1.18
90-93
67
1
0.38
5.23
1.33
2
0.25
4.62
1.00
3
0.24
4.21
0.94
IM240-Ph 2
0.22
4.10
0.92
1994+
66
1
0.13
1.50
0.46
2
0.05
1.06
0.35
3
0.05
1.00
0.33
IM240-Ph 2
0.04
0.95
0.32
LDT
81-85
12
1
1.98
23.47
4.59
2
1.51
22.20
4.26
3
1.36
20.90
4.09
IM240-Ph 2
1.31
21.14
3.98
86-89
23
1
1.37
18.43
2.29
2
0.97
16.41
2.06
3
0.84
15.59
2.00
IM240-Ph 2
0.85
15.08
2.02
90-93
41
1
0.51
6.74
1.58
2
0.28
5.19
1.22
3
0.25
4.38
1.21
IM240-Ph 2
0.24
5.24
1.10
1994+
35
1
0.25
3.77
1.05
2
0.15
2.59
0.83
3
0.14
2.44
0.78
IM240-Ph 2
0.13
2.86
0.72
In addition to the triplicate IM147 emission rates (presented separately for test sequence 1
to 3), Table 3-1 contains the mean score for the end portion of the IM240 test that was
conducted after the third IM147 test.
Substantial reductions in mean IM147 scores are observed between the first and second
test in the sequence, with lesser reductions occurring between the second and third test in
the sequence. For most model year groups and pollutants, emissions in the end portion of
the IM240 continue to decline, although the reductions are diminished relative to those
-19-
-------
observed during the triplicate IM147 testing. In some cases, the end portion of the IM240
shows an increase in emissions relative to the third IM147. This is most pronounced for
the 1986 to 1989 model year passenger car group, which shows an increase of 11% for
HC and an increase of 28% for CO between the third IM147 and the end portion of the
IM240. However, as discussed below, this effect is primarily the result of two vehicles
that had much higher HC and CO emissions in the IM240 portion than in the IM147
portions of the test sequence.
Comparison of Failure Rates
As part of the recent DEQ project, Sierra developed alternative IM240 and IM147
cutpoints that resulted in similar failure rates.6 Consistent with the requirements of that
project, the "Alternative #2" IM240 cutpoints selected in the DEQ study were designed to
maximize potential CO benefits while keeping failure rates (particularly for older model
year vehicles) at an acceptable level (e.g., an approximate 50% failure rate for 1981 to
1985 model year vehicles). These cutpoints generally fall between the EPA start-up and
final cutpoints, but in some cases (particularly for the light truck categories) the CO
cutpoints are more stringent than the final EPA cutpoints.
The cutpoints developed for the DEQ project were used with the current test results to
make a first-cut comparison in failure rates between IM147 testing and full IM240
testing. A summary of those cutpoints is given in Tables 3-2 and 3-3 for full IM240
testing and IM147 testing, respectively.
As reflected in Table 3-2, the full IM240 procedure incorporates a "two-ways-to-pass"
methodology (i.e., separate cutpoints apply to the end portion of the IM240 or IM147,
and if a vehicle passes the EM 147, it passes the test). This approach allows for the first
part of the test to serve as a preconditioning cycle for cases in which the IM147 emissions
are below the IM147 cutpoints. In our previous analysis of IM147-only testing, it was
necessary to also develop criteria for a two-ways-to-pass approach for that test. This was
accomplished by establishing IM147 Phase 1 and Phase 2 components of the test and
developing separate cutpoints for Phase 2. The test cycle is illustrated in Figure 3-3.
Cutpoints for Phase 2 were generated through a regression analysis relating Phase 2 to the
entire IM147 emission rates. These correlation equations were then applied to the full
IM147 cutpoints in Table 3-3. For this evaluation, the Phase 2 cutpoints were developed
in terms of total mass of HC, CO, or NOx in Phase 2, but g/mi cutpoints could also be
established if desired.
The following Phase 2 correlation equations were used to determine the Phase 2
cutpoints:
HC phase 2
0.880 x HC [M 147 - 0.002
(3-1)
CO Phase 2
0.934 x CO IM ,47 - 0.689
(3-2)
1.034 x NOx IM147- 0.064
(3-3)
-20-
-------
Table 3-2
DEQ "Alternative #2" Full-Cycle IM240 Cutpoints
(Composite/IM147 Cutpoints in g/mi)
Vehicle Class
Model Years
HC
CO
NOxa
LDGV
1981-82
3.00/1.88
25.0/20.0
3.0
1983-85
2.20/1.38
16.0/12.8
3.0
1986-89
1.50/0.94
15.0/12.0
2.5
1990-93
1.00/0.63
12.0/9.6
2.5
1994+
0.80/0.50
12.0/9.6
2.0
LDGT1
1981-85
4.00/2.50
30.0/24.0
5.0
1986-89
3.00/1.88
25.0/20.0
4.5
1990-93
2.00/1.25
20.0/16.0
4.0
1994+
1.60/1.00
20.0/16.0
3.0
LDGT2
1981-85
4.40/2.75
50.0/40.0
6.0
1986-87
4.00/2.50
40.0/32.0
5.5
1988-89
3.00/1.88
25.0/20.0
5.5
1990-93
3.00/1.88
25.0/20.0
5.0
1994+
2.40/1.50
25.0/20.0
4.0
a NOx IM147 cutpoints are equal to the NOx composite cutpoints.
-21-
-------
Table 3-3
DEQ "Alternative #2" IM147 Cutpoints (g/mi)
Vehicle Class
Model Years
HC
CO
NOx
LDGV
1981-82
3.0
25
3.5
1983-85
2.4
20
3.5
1986-89
1.6
15
2.5
1990-93
1.0
12
2.5
1994+
0.8
12
2.0
LDGT1
1981-85
4.0
40
5.5
1986-89
3.0
25
4.5
1990-93
2.0
20
4.0
1994+
1.6
20
3.0
LDGT2
1981-85
4.4
48
7.0
1986-87
4.0
40
5.5
1988-89
3.0
25
5.5
1990-93
3.0
25
5.0
1994+
2.4
25
4.0
Thus, the Phase 2 cutpoint for a IM147 HC cutpoint of 0.8 g/mi would simply be:
HC Phase 2 = (0.880 x 0.8 g/mi)-0.002= 0.702 grams (3-4)
-22-
-------
Figure 3-3
IM147 Trace Showing Phases 1 and 2
Time (sec)
Using the outpoints in Tables 3-2 and 3-3 in conjunction with the test scores from the
triplicate EM 147 testing, failure rates were calculated for each IM147 and for the full
IM240. The results of this analysis are presented in Table 3-4. As observed in that table,
the failure rates for the IM147 testing generally drop from the first to the second test.
Table 3-4
IM147 and IM240 Failure Rates
Based on DEQ "Alternative #2" Cutpoints
Model
Year
Group
IM147 Failure Rates
IM240
Failure
Rates
First
Second
Third
1981 - 1985
31.3%
31.3%
31.3%
37.5%
1986- 1989
20.6%
15.9%
17.5%
19.1%
1990- 1993
11.1%
8.3%
7.4%
8.3%
1994+
1.0%
1.0%
1.0%
2.0%
All Vehicles
11.8%
9.9%
9.9%
11.5%
-23-
-------
Between the second and the third IM147, the failure rate for the 1986 - 1989 model year
group actually increases. For the most part, this is a result of vehicles being just under
the cutpoints for the second test and just over the cutpoint for the third test. Comparing
the full IM240 failure rates to the IM147 failure rates, one observes that the full IM240
failure rates are slightly higher relative to the second and third IM147 tests. (This is the
fairest comparison, since the first IM147 test represents vehicles in a different state of
warm-up.) Inspection of the test results indicates that this is a result of imperfections in
developing the IM147 cutpoints for the DEQ project. A number of the vehicles that
failed the full IM240 test and passed the second or third IM147 test were just over the
IM147 cutpoint, either for the full IM147 or Phase 2.
Of particular interest in the analysis of data from the triplicate IM147 testing are the
vehicles that had inconsistent pass/fail scores across the series of four tests. These
vehicles are highlighted in Table 3-5, which shows HC, CO, and NOx emissions results
and the pass/fail status (based on the DEQ Alternative #2 cutpoints) for each test. In
total, there were 24 vehicles that fell into this group. Eight of those vehicles failed the
first IM147 or the first and second IM147 and then went on to pass the remaining tests in
the series (i.e., record numbers 1,6, 11, 12, 16, 20, 21, and 22 in Table 3-5). It is likely
that these vehicles were not sufficiently preconditioned at the start of testing. (Later in
this report, the development of an algorithm to identify such vehicles as needing a retest
is described.)
Several other vehicles in Table 3-5 are very close to the cutpoints and, after passing the
first test, went on to fail the second and third IM147 as well as the full IM240 test (e.g.,
record numbers 2 and 4). One of these vehicles (number 2) appears to have passed the
Phase 2 portion of the test on the first test, but failed the remaining tests. This indicates
that it may be appropriate to build in a safety margin in the Phase 2 cutpoints. Consistent
with the approach used with the IM240 test, that approach was considered in the full set
of IM147 cutpoints developed later in this effort. Finally, there are a few vehicles that
passed all three IM147 tests but failed the full IM240. This points to an imperfection in
matching failure rates for the IM147 and the full IM240.
There are four vehicles in Table 3-5 that stand out: record numbers 14, 15, 23, and 24.
All of these vehicles had significantly higher HC and CO emissions during the full
IM240 procedure than during the first three IM147 tests. To better understand what is
happening to these vehicles, it is instructive to review the second-by-second results from
the series of tests. The results for CO are shown in Figures 3-4, 3-5, 3-6, and 3-7 for
record numbers 14, 15, 23, and 24, respectively. Below is a brief discussion of each
vehicle.
• Record 14, a 1988 Pontiac Bonneville, had relatively low CO emissions during
the first and second IM147 test (1.35 and 3.42 g/mi, respectively). However, CO
emissions during the third IM147 increased substantially (to 55.72 g/mi) and were
higher still during the IM240 following the IM147 testing. It is interesting to note
that CO was emitted in measurable quantities throughout the test, and the large
increases are not attributable to a specific section of the trace. It thus appears that
the gradual emissions increase could be attributable to excessive purge as the
vehicle warmed up or to some kind of catalyst protection scheme.
-24-
-------
Table 3-5
Summary of Vehicles in the Triplicate IM147 Test Program with Non-Matching Pass/Fail Results
Based on ADEQ Alternative #2 Outpoints
Record
Number
Model
Year
Veh
Type
First IM147
Second IM147
Third IM147
Full IM240 - Com
posite and IM147 Scores
HC
CO
NOx
P/F
HC
CO
NOx
P/F
HC
CO
NOx
P/F
CHC
CCO
CNOX
P2HC
P2CO
P2NOX
P/F
1
1981
LDGV
1.68
15.17
3.89
F
1.59
14.26
3.55
P
1.27
12.76
3.28
P
0.86
11.38
2.89
0.91
11.02
3.15
P
2
1984
LDGV
1.63
20.63
1.94
P
1.51
20.24
1.85
F
1.46
24.42
1.72
F
1.46
25.76
1.46
1.46
26.90
1.62
F
3
1984
LDGT1
2.51
43.33
2.64
P
2.56
41.09
2.31
P
2.35
41.42
2.58
P
3.38
59.47
1.70
2.54
48.78
2.11
F
4
1984
LDGT1
3.49
37.04
3.91
P .
3.10
43.21
3.40
F
3.13
45.33
3.31
F
2.79
42.40
3.00
2.96
51.42
3.06
F
5
1985
LDGV
1.04
17.69
0.81
P
0.81
18.60
0.62
P
0.74
17.64
0.67
P
0.88
20.15
0.60
0.99
19.76
0.71
F
6
1985
LDGT2
1.56
1.70
7.68
F
1.04
1.22
5.25
P
0.77
1.16
2.86
P
0.74
0.84
2.35
0.69
0.64
2.82
P
7
1986
LDGV
2.12
29.95
0.80
F
1.03
17.54
0.62
P
0.70
17.62
0.55
F
0.69
16.56
0.36
0.60
15.58
0.36
F
8
1986
LDGT1
1.55
20.32
1.08
P
1.27
21.03
0.81
P
1.39
25.86
1.12
F
1.68
26.86
1.25
1.63
26.99
1.48
F
9
1987
LDGV
0.21
3.70
2.65
P
0.41
12.66
3.04
F
0.07
2.42
2.22
P
0.04
0.94
1.77
0.05
1.21
1.94
P
10
1987
LDGV
0.34
1.02
2.93
F
0.24
1.07
2.73
F
0.21
0.85
2.67
F
0.21
0.95
2.45
0.20
0.92
2.68
P
11
1987
LDGV
0.69
15.71
1.61
F
0.56
8.98
1.75
P
0.56
13.40
1.54
P
0.44
8.46
1.47
0.49
9.41
1.70
P
12
1987
LDGV
0.93
5.47
3.19
F
0.20
1.57
1.54
P
0.13
1.21
1.56
P
0.12
0.95
1.61
0.13
1.10
1.51
P
13
1987
LDGT2
4.70
22.63
7.31
F
3.43
16.98
7.68
F
2.01
10.54
5.16
P
2.51
14.22
6.09
2.60
13.94
6.73
F
14
1988
LDGV
0.13
1.35
0.33
P
0.14
3.42
0.19
P
1.49
55.72
0.58
F
2.36
96.25
0.35
2.16
88.69
0.47
F
15
1989
LDGV
0.47
15.03
2.15
F
0.76
17.82
1.88
P
0.45
14.37
1.70
P
2.64
83.25
0.76
3.35
105.61
0.65
F
16
1990
LDGV
0.49
20.61
0.30
F
0.48
13.33
0.31
P
0.42
10.57
0.44
P
0.29
11.51
0.35
0.36
11.32
0.35
P
17
1991
LDGV
0.87
6.53
3.23
F
0.87
6.84
2.81
F
0.79
7.23
2.78
F
0.76
6.85
2.43
0.77
6.85
2.70
P
18
1991
LDGV
0.49
5.95
1.74
P
0.33
9.66
1.04
P
0.31
10.04
0.97
P
0.48
13.82
0.94
0.46
11.62
1.00
F
19
1992
LDGV
1.66
15.26
1.57
F
0.87
8.24
1.31
P
1.42
10.74
1.25
F
1.10
8.27
1.13
1.01
10.84
1.28
F
20
1992
LDGT1
0.65
20.32
1.09
F
0.57
26.21
0.99
F
0.34
17.03
0.90
P
0.23
13.06
0.69
0.31
16.99
0.82
P
21
1993
LDGV
0.64
18.26
1.57
F
0.35
11.28
1.52
P
0.44
12.38
1.60
P
0.13
7.68
1.52
0.10
7.13
1.70
P
22
1993
LDGV
0.72
7.27
3.84
F
0.04
0.28
2.70
F
0.01
0.56
1.80
P
0.01
0.29
2.11
0.00
0.16
2.85
P
23
1993
LDGT1
0.15
10.80
1.60
P
0.10
8.29
1.30
P
0.10
9.16
0.99
P
0.39
32.09
0.62
0.54
44.29
0.64
F
• 24
1995
LDGT1
0.08
2.72
0.51
P
0.05
0.84
0.35
P
0.04
0.72
0.44
P
0.40
20.28
0.42
0.55
27.99
0.49
F
-------
Figure 3-4
Second-by-Second CO Emissions from Triplicate IM147 Testing
1988 Pontiac Bonneville (Record No. 14)
First IM147 (1.35 g/mi)
Target Speed jT
\
j/ Vehicle Speed
\
y Emissions
\
,/ ..... ,.
0 20 40 60 80 100 120 140
Time (sec)
Time (sec)
IM240 CO Emissions
-26-
-------
Figure 3-5
Second-by-Second CO Emissions from Triplicate IM147 Testing
1989 Dodge Dynasty (Record No. 15)
0 20 40 60 80 100 120 140
Time (sec)
Time (sec)
IM240 CO Emissions
-27-
-------
Figure 3-6
Second-by-Second CO Emissions from Triplicate IM147 Testing
1993 Ford Ranger (Record No. 23)
First IM147 (10.80 g/mi)
0 20 40 60 80 100 120 140
Time (sec)
Time (sec)
-28-
-------
Figure 3-7
Second-by-Second CO Emissions from Triplicate IM147 Testing
1995 Toyota 4Runner (Record No. 24)
-29-
-------
Record 15, a 1989 Dodge Dynasty, had moderate CO emissions during the first
three IM147 tests (14 to 18 g/mi), but emissions during the IM240 test were
excessive, particularly during the end portion of that test (106 g/mi). Reviewing
the modal CO emissions in Figure 3-5, one observes that the vehicle appears to go
into open-loop operation at the start of the large hill of the end portion of the test
(i.e., beginning at about second 160 of the IM240). Although CO emissions
accrue throughout this test, the period from 160 to 230 comprises the bulk of the
emissions.
• Record 23, a 1993 Ford Ranger, shows a very similar emissions response
throughout the three IM147 tests. As seen in Figure 3-6, most of the CO
emissions occur during seconds 62 to 75 of the IM147. During the end portion of
the IM240, a similar pattern is observed. In that test, however, substantial CO is
also emitted during the high-speed portion of the trace. It is not entirely clear
what has caused this, but it appears that the vehicle did not follow the speed-time
trace as smoothly during the end portion of the IM240 as it did during the first
three IM147 tests.
Record 24, a 1995 Toyota 4Runner, had decreasing emissions throughout the first
three IM147 tests, emitting only 0.72 g/mi CO during the third IM147.
However, the composite IM240 CO emission rate for this vehicle was 20.28 g/mi
and the emissions reading during the end portion of the IM240 was 27.99 g/mi.
As observed in Figure 3-7, the bulk of the emissions for this test occur during
seconds 160 to 180 where it appears that the vehicle went into open-loop
operation. Overall, this vehicle did not follow the speed-time profile as closely as
the vehicles shown in Figures 3-4 to 3-6.
Three of the four vehicles identified above had similar emissions characteristics in that
the bulk of the emissions during the IM240 was related to apparent open-loop operation
during part of the end portion of the test that was not observed during the first three
IM147 tests. Record 14 was unusual in that CO emissions continually increased
throughout the series of tests. Overall, the addition of the Phase 1 portion of the IM240 ,
did little in terms of contributing to the high emissions observed in the IM240.
It was not possible to satisfactorily resolve what is happening with these vehicles.
Because the only IM240 that is used for comparison is at the end of the testing, the
analysis of excess emissions identified by the IM147 test that uses the final IM240 as the
"gold standard" will reflect poorly on IM147 testing since it would falsely pass several
vehicles with very high CO emissions. (Just considering the 1988 Bonneville and the
1989 Dynasty above would reduce the effectiveness of the EM147 procedure to
unacceptable levels.) However, it is apparent that the higher emissions certain vehicles
exhibited during the IM240 are due to some anomaly and not to a fundamental problem
with the IM147 trace. For these reasons, test results for the four vehicles described above
are considered outliers and were not used in the remaining analyses of IM147 testing.
###
-30-
-------
4. IM147 CUTPOINT ANALYSIS
This section of the report presents the results of Sierra's development of IM147 outpoints,
development of fast pass criteria, and development of predictive criteria that can be used
to determine if a retest is warranted for a failure. In addition, the effect of exempting up
to the first 10 model years of vehicles from testing was evaluated.
The following three sets of cutpoints were developed:
• Startup cutpoints;
• Intermediate cutpoints;
• Final cutpoints.
For each cutpoint, both overall and "two ways to pass" ("Phase 2") standards were
developed. For each set of standards, an assessment of the excess emission losses using
the IM147 versus the IM240 was performed, and the potential effect on test time from
using the IM147 was evaluated. The test time modeling was performed with the sample
test fleet data using the IM147 (with a maximum of two retests) versus using the IM240
(with a maximum of one retest).
Next, methods to reduce the test time were evaluated. If test time can be reduced, this is
equivalent to increasing the testing network's capacity at no additional cost. In addition,
reductions in test time can also help to reduce wait times for consumers. Previous studies
by Sierra have investigated several techniques that can be applied to the test data from
vehicles while they are being tested to reduce the overall test time with only minor losses
in the excess emissions identified. The first method developed was fast-pass standards to
allow very clean vehicles to exit the test early. The next method developed was retest
algorithms that predict if a failing vehicle would pass if retested. If a vehicle would not
pass if retested, then retesting the vehicle is not warranted (since it would be increasing
the test time with no benefit). The last technique evaluated was model year exemptions.
Exempting newer model year vehicles from testing for their first four or five years can
significantly reduce total test time for the fleet. Since new vehicles rarely fail I/M tests,
the excess emissions lost from not testing these vehicles should be small.
In this section of the report, the methodology used to develop the standards is discussed,
and the effects on test time and excess emissions identified for the IM147 and the IM240
are compared. The development of fast pass cutpoint coefficients that allow the
prediction of full test results from various designated modes on the test are then
presented, followed by the retest algorithms and model year exemptions. Each of these
-31-
-------
techniques to reduce test time was also evaluated for its impact on reducing test time at
the cost of potentially losing some excess emissions. Testing of each technique built on
the previous technique; i.e., testing of the retest algorithms was done while first applying
the fast pass criteria. It is important for the reader to keep this in mind because there are
several instances where this building-block approach produces unexpected but beneficial
results.
Cutpoint Selection
Development of the cutpoints for the IM147-only test was based on linear regressions
between full (composite) IM240 emission rates and emission rates for the end portion of
the same IM240 tests.* The IM240 test data used were from the 304 vehicles described in
Section 3, with the four outliers removed from the data set. (As noted in Section 3, these
vehicles were removed from all further analyses.) These data were used because the
IM240 tests were conducted after the three IM147 tests, so all vehicles should have been
adequately warmed up and there should have been no preconditioning effects. Linear
regressions were performed for each model of four year groups. The results are shown in
Table 4-1.
Table 4-1
Regression Coefficients of IM240 Versus IM147 Emission Rates
1981 - 85
1986 - 89
1990 - 1993
1994 +
Slope
(m)
y-lntercept
(b)
Slope
(m)
y-
Intercept
(b)
Slope
(m)
y-
Intercept
(b)
Slope
(m)
y-Intercept
(b)
HC
0.884
0.110
1.05
0.00968
1.09
0.00242
0.994
0.0151
CO
0.928
1.86
1.07
0.271
1.02
0.106
1.02
0.152
NOx
1.06
0.156
0.976
0.185
0.106
0.0216
1.08
0.0408
The regression equations were applied to the composite IM240 emissions standards using
the following equation to develop the IM147 composite emissions standards.
IM147 standard (g/mi) = (m x IM240 cutpoint (g/mi) ) +b (4-1)
where: m = slope of regression equation
b = y intercept of regression equation
#
The IM147 portion of the same IM240 was used to develop initial sets of cutpoints to ensure an equal
level of preconditioning. If the cutpoints were instead developed using the first IM147 versus the IM240
that was conducted after the triplicate IM147s, it is likely that preconditioning issues would have clouded
the results. Use of the first IM147 for the correlation analysis would also have introduced a difficulty in
how to deal with the second and third IM147s in the series.
-32-
-------
The correlation results were used as a general guide regarding where to set the outpoints,
with minor subsequent adjustments being made to better match IM240 and IM147 failure
rates. In some cases, use of these regressions yielded cutpoints that were discontinuous
(e.g., some newer model years yielded higher emissions rates than older years). This was
because the relationships between the composite IM240 and the IM147 emission rates
were developed using real data in which the ratio of emissions varied by model year. In
these cases, the cutpoints were modified to decrease with newer model years. The
cutpoint categories were then modified to match as close as possible the EPA model year
categories, but in some cases a few new categories were added due to significant changes
between model years.
Using the regressions and then modifying the resulting cutpoint as described above to
account for discontinuities, startup, intermediate, and final cutpoints were developed for
LDGVs, LDGTls and LDGT2s. The intermediate cutpoints were based simply on
applying the regression equations to the average of the IM240 composite startup and the
final cutpoints. The resulting standards are shown in Table 4-2 through Table 4-4 for
startup standards, intermediate standards, and final standards, respectively.
To evaluate the impact of these new cutpoints on test time and excess emissions
identified, the data from the 300 vehicle data set used to develop the standards were
"tested" using the two-way-to-pass methodology for all three IM147s (i.e., if a vehicle
failed the first IM147, the data from the second test were evaluated to see if the vehicle
would pass or fail; if it failed the second test, the data from the third test were evaluated).
The IM240 at the end of the testing pattern was used as the standard for comparing excess
emissions identified, with excess emissions defined as emissions above the cutpoints on
the IM240. If a vehicle failed the IM240 and the vehicle also failed the IM147, then the
IM147 would be attributed with identifying the excess emissions from that vehicle, and
those emissions would have a value equal to the IM240 emissions over the IM240
standard. If the IM147 did not fail a vehicle that failed the IM240, the excess emissions
from the IM240 (emissions over the IM240 standard for the IM240 test) were considered
excess emissions not identified by the IM147. The results of the analysis are presented in
Table 4-5.
The ratio of excess emissions for vehicles that failed the IM147 versus those failing the
IM240 is the percent excess emissions identified.
As can be seen in the Table 4-5, there were few excess emissions "lost" (i.e., the IM147
failed almost all of the vehicles that the IM240 failed). For startup standards, the IM147
identified all of the excess CO and NOx emissions the IM240 identified and almost all of
the HC. Only 1.5% of the excess HC emissions were lost with the intermediate standards
and 6.8% of the NOx. The maximum loss using the final standards was 2.5% of excess
CO emissions.
To more objectively evaluate their impact, it would be preferable to test the cutpoints on a
separate data set. However, the only such sample is the 101 vehicles previously tested in
Arizona using triplicate IM147s after a 30-minute idle. These data are not considered
representative of average IM240 testing in Arizona, which typically involves wait times
-33-
-------
Table 4-2
Startup IM240 Cutpoints and IM147 Cutpoints Developed in This Study
(Composite/Phase 2 Cutpoints in g/mi, IM240 - IM147)
Model Years
HC
CO
NOxa
LDGV
1981-82
2.00/1.25 -2.00/1.20
60.0/48.0-58.0/30.0,
3.0-3.3/1.2
1983-85
2.00/1.25 -2.00/1.20
30.0/24.0-30.0/15.0
3.0-3.3/1.2
1986-90
2.00/1.25 -2.00/1.00
30.0/24.0 - 30.0/10.0
3.0-3.0/1.2
1991-93
1.20/0.75 - 1.30/0.60
20.0/16.0-21.0/10.0
2.5-2.9/1.0
1994-95
1.20/0.75 - 1.20/0.60
20.0/16.0-21.0/10.0
2.5-2.7/1.0
1996+ (Tier l)b
0.80/0.50 - 0.80/0.50
15.0/12.0 - 15.0/ 7.0
2.0-2.1/0.9
LDGT1
1981-83
7.50/5.00 - 6.70/4.70
100.0/80.0 - 95.0/50.0
7.0 - 7.6/2.9
1984-85
3.20/2.00 - 2.90/2.00
80.0/64.0 - 76.0/40.0
7.0 - 7.6/2.9
1986-87
3.20/2.00-2.90/1.60
80.0/64.0-76.0/31.0
7.0 - 7.0/2.7
1988-90
3.20/2.00-2.90/1.60
80.0/64.0-76.0/31.0
3.5-3.6/1.3
1991-93
2.40/1.50-2.60/1.20
60.0/48.0-61.0/31.0
3.0-3.4/1.1
1994-95
2.40/1.50-2.40/1.20
60.0/48.0-61.0/29.0
3.0-3.2/1.1
1996+ (Tier l)b'c
1.00/0.63 - 1.0/0.60
20.0/16.0-21.0/10.0
2.5-2.7/1.1
LDGT2
1981-83
7.50/5.00 - 6.70/4.70
100.0/80.0-95.0/50.0
7.0 - 7.6/2.9
1984-86
3.20/2.00 - 2.90/2.00
80.0/64.0 - 76.0/40.0
7.0 - 7.6/2.9
1987
3.20/2.00-2.90/1.60
80.0/64.0-76.0/31.0
7.0 - 7.6/2.7
1988-90
3.20/2.00-2.90/1.60
80.0/64.0- 76.0/31.0
t-ft
0
1
»—*
VO
1991-93
2.40/1.50- 2.60/1.20
60.0/48.0-61.0/31.0
4.5-5.1/1.9
1994-95
2.40/1.50- 2.40/1.20
60.0/48.0-61.0/29.0
°)
•—.
oo
1
in
1996+ (Tier l)b'c
2.40/1.50-2.40/1.20
60.0/48.0-61.0/29.0
4.0-4.3/1.7
a NOx Phase 2 cutpoints are equal to the NOx composite outpoints for the IM240.
b Because there is no way to discern Tier 1 and Tier 0 vehicles in the Arizona database, the Tier 1 IM240
cutpoints were not implemented until model year 1996, when all vehicles must certify to the Tier 1 standards.
To be conservative, the Tier 1 LDGT1 and LDGT2 cutpoints used for this analysis reflect the cutpoints
recommended by EPA for the heavier vehicles in each class.
-34-
-------
Table 4-3
Intermediate IM240 Cutpoints and IM147 Cutpoints Developed in This Study
(Composite/Phase 2 Cutpoints in g/mi, IM240 - IM147)
Model Years
HC
GO
NOx"
LDGV
1981-82
1.40/0.88- 1.40/0.90
45.0/36.0-44.0/23.0
2.3-2.8/1.0
1983-85
1.40/0.88- 1.40/0.90
23.0/18.0-23.0/12.0
2.3-2.8/1.0
1986-90
1.40/0.88- 1.40/0.70
23.0/18.0-23.0/9.0
2.3-2.6/1.0
1991-93
1.00/0.63 - 1.10/0.50
18.0/14.0-18.0/9.0
2.3 - 2.6/0.9
1994-95
1.00/0.63 - 1.00/0.50
18.0/14.0-18.0/9.0
2.3 - 2.5/0.9
1996+ (Tier l)b
0.70/0.45 - 0.80/0.40
13.0/10.0- 15.0/6.0
1.8-2.2/0.8
LDGT1
1981-83
5.50/3.50-4.90/3.40
85.0/68.0- 81.0/43.0
5.8 - 6.3/2.4
1984-85
2.40/1.50 - 2.30/1.50
60.0/48.0 - 60.0/30.0
5.8 - 6.3/2.4
1986-87
2.40/1.50-2.30/1.20
60.0/48.0 - 60.0/26.0
5.8 - 5.8/2.2
1988-89
2.40/1.50 - 2.30/1.20
60.0/48.0 - 60.0/26.0
3.0-3.3/1.2
1990
2.40/1.50 - 2.30/1.20
60.0/48.0 - 59.0/26.0
3.0-3.3/1.2
1991-93
2.00/1.25 -2.10/1.00
50.0/40.0 - 51.0/26.0
2.8-3.2/1.1
1994-95
2.00/1.25 -2.00/1.00
50.0/40.0- 51.0/25.0
2.8-3.0/1.1
1996+ (Tier l)bc
0.90/0.57- 1.30/0.60
17.0/13.0-31.0/8.0
2.2-2.7/1.0
LDGT2
1981-83
5.50/3.50 - 4.90/3.40
85.0/68.0- 81.0/43.0
5.8 - 6.3/2.4
1984-86
2.40/1.50-2.30/1.50
60.0/48.0 - 60.0/30.0
5.8 - 6.3/2.4
1987
2.40/1.50-2.30/1.20
60.0/48.0 - 60.0/26.0
5.8 - 6.3/2.2
1988-90
2.00/1.50-2.30/1.20
60.0/48.0 - 60.0/26.0
4.3-4.6/1.6
1991-93
2.00/1.25 -2.20/1.00
50.0/40.0- 51.0/26.0
4.0-4.6/1.6
1994-95
2.00/1.25 -2.00/1.00
50.0/40.0-51.0/25.0
4.0-4.3/1.6
1996+ (Tier l)bc
1.60/1.00-2.00/0.90
38.0/30.0-51.0/18.0
3.0-4.1/1.3
a NOx Phase 2 cutpoints are equal to the NOx composite cutpoints for the IM240.
k Because there is no way to discern Tier 1 and Tier 0 vehicles in the Arizona database, the Tier 1 IM240
cutpoints were not implemented until model year 1996, when all vehicles must certify to the Tier 1 standards.
c To be conservative, the Tier 1 LDGT1 and LDGT2 cutpoints used for this analysis reflect the cutpoints
recommended by EPA for the heavier vehicles in each class.
-35-
-------
Table 4-4
Final IM240 Cutpoints and IM147 Cutpoints Developed in This Study
(Composite/Phase 2 Cutpoints in g/mi, IM240 - IM147)
Model Years
HC
CO
NOxa
LDGV
1981-1982
0.80/0.50 - 0.80/0.50
30.0/24.0- 30.0/15.0
2.0 - 2.3/0.8
1983-1985
0.80/0.50 - 0.80/0.50
15.0/12.0- 16.0/8.0
2.0 - 2.3/0.8
1986-1989
0.80/0.50 - 0.80/0.40
15.0/12.0- 16.0/8.0
2.0 - 2.2/0.8
1990-1993
0.80/0.50 r 0.80/0.50
15.0/12.0- 15.0/8.0
2.0 - 2.2/0.7
1994-1995*
0.80/0.50 - 0.80/0.50
15.0/12.0-15.0/7.0
2.0 - 2.2/0.7
1996+ (Tier l)b
0.60/0.40 -0.80/0.30
10.0/8.0- 15.0/5.0
1.5-2.2/0.6
LDGT1
1981-83
3.40/2.00- 3.10/2.10
70.0/56.0 - 67.0/35.0
4.5-4.9/1.8
.1984-85
1.60/1.00- 1.70/1.00
40.0/32.0-43.0/20.0
4.5-4.9/1.8
1986-87
1.60/1.00- 1.70/0.80
40.0/32.0 - 43.0/20.0
4.5-4.6/1.7
1988-1989
1.60/1.00- 1.70/0.80
40.0/32.0 -43.0/20.0
2.5-2.9/1.0
1990-1993
1.60/1.00- 1.60/0.80
40.0/32.0 -41.0/20.0
2.5-2.9/1.0
1994-1995
1.60/1.00- 1.60/0.80
40.0/32.0-41.0/20.0
2.5-2.7/1.0
1996+ (Tier l)bc
0.80/0.50- 1.60/0.50
13.0/10.0-41.0/6.0
1.8-2.7/0.8
LDGT2
1981-83
3.40/2.00- 3.10/2.10
70.0/56.0 - 67.0/35.0
4.5-4.9/1.8
1984-86
1.60/1.00- 1.70/1.00
40.0/32.0 - 43.0/20.0
4.5-4.9/1.8
1987
1.60/1.00 - 1.70/0.80
. 40.0/32.0 -43.0/20.0
4.5-4.9/1.7
1988-1991
1.60/1.00- 1.70/0.80
40.0/32.0 -43.0/20.0
3.5-4.0/1.3
1992-1993
1.60/1.00- 1.70/0.80
40.0/32.0 -41.0/20.0
3.5-4.0/1.3
1994-1995
1.60/1.00- 1.70/0.80
40.0/32.0-41.0/20.0
3.5 - 3.8/1.3
1996+ (Tier l)b c
0.80/0.50- 1.60/0.50
15.0/12.0-41.0/7.0
2.0 - 3.8/0.9
a NOx Phase 2 cutpoints are equal to the NOx composite cutpoints for the IM240.
b Because there is no way to discern Tier 1 and Tier 0 vehicles in the Arizona database, the Tier 1 IM240
cutpoints were not implemented until model year 1996, when all vehicles must certify to the Tier 1 standards.
To be conservative, the Tier 1 LDGT1 and LDGT2 cutpoints used for this analysis reflect the cutpoints
recommended by EPA for the heavier vehicles in each class. .
-------
Table 4-5
Evaluation of Emissions Standards for the IM147
Impact on Test Time and Excess Emissions Lost
Vehicle
Class
Model
Year
Group
IM147
Number in Test
Sample
Mean Test
Time (seconds)
% Excess Emissions Identified"
HC
CO
NOx
Startup Outpoints
LDGV
81-84
13
294
100
100
100
85-89
45
180
100
100
-
1990+
133
159
100
100
100
LDGT
1&2
81-84
8
239
93.8
-
100
85-89
27
185
100
100
-
1990+
74
159
-
-
100
Sum / Weighted average
300
172
97.2
100
100
Intermediate Cutpoints
LDGV
81-84
13
328
100
100
98.1
85-89
45
209
100
100
98.1
1990+
133 .
166
95.3
100
100
LDGT
1&2
81 - 84
8
294
100
100
76.1
85-89
27
201
90.3
100
0
1990+
74
163
100
-
100
Sum / Weighted average
300
185
98.5
100
93.2
Final Cutpoints
LDGV
81-84
13
362
100
100
100
85-89
45
255
98.3
100
95.9
1990+
133
171
100
100
96.4
LDGT
1&2
81 - 84
8.
368
100
100
100
85-89
27
289
97.2
89.2
100
1990+
74
165
100
-
100
Sum / Weighted average
300
206
99.2
97.5
99.4
a indicates there were no failures in this group for the IM240 and therefore no excess emissions for
the IM147 to identify.
-37-
-------
of no more than 5-15 minutes. For this reason, no evaluation of the outpoints was
performed using the previously collected data.
The results of the third IM147 test versus the IM240 were the same for 295 of the 300
vehicles that were tested using the final standards. The third IM147 was used for this
comparison because it was assumed the vehicle was completely warmed up at this point
in the testing. A total of only 45 vehicles (15%) failed either the IM240 or the IM147
cutpoints, with 40 of these vehicles failing both the IM240 and IM147 cutpoints. There
are five vehicles for which the results did not agree: three that failed the IM240 and
passed the IM147 cutpoints (errors of omission, the cause of the excess emissions not
identified); and two that passed the IM240 and failed the IM147 cutpoints (errors of
commission, with the IM147 having higher excess emissions identified than the IM240).
The data for each of these five vehicles are presented in Table 4-6 and discussed in detail
below.
The first two vehicles in Table 4-6 failed the IM240 but did not fail the IM147 cutpoints.
These are referred to as Type I errors in the table.
• The first vehicle was a LDGV that failed for NOx, and was approximately 5%
over the IM240 cutpoint but 18% below the IM147 cutpoint for the third test.
The NOx emissions from the first three tests were dropping quickly and the
vehicle did not pass the IM147cutpoints until the last test. This vehicle may
have been fairly cold at the beginning of the testing, but was warming up as
the testing continued.
• The second vehicle was a LDGT1 that failed for HC and was also 5% over the
IM240 cutpoint, but had been below the IM147 cutpoint for all three tests.
However, this vehicle showed a decrease in emissions from the first to the
second test, then an increase to the third of the three EM 147 tests. In fact, the
emissions of all three pollutants increased from the second to the third test,
indicating that the vehicle may have had some sort of transient emissions
control problem that led to enough variability in the vehicle's emissions to
cause differing results.
The last three vehicles in Table 4-6 are all LDGVs and passed the IM240 but failed the
IM147 cutpoints. These are referred to as Type II errors in the table.
• The first vehicle failed the IM147 CO cutpoint by 0.01 g/mi, which is less
than the probable sampling error in the measurement. In addition, if the test
lane software rounds or truncates the result, the vehicle could have passed.
The data from the three IM147 tests show the emissions rate between the three
tests first went down and then back up (the vehicle first failed the IM147
cutpoint, then passed, then failed). In actual testing, this vehicle would have
passed the second test and not been tested again (the third time), giving the
same result as the IM240.
-38-
-------
Table 4-6
Tests with Type 1 and Type II Errors (IM240 and Third IM147 Test Results Differ)
IM240 Fail - Third IM147 Test Pass (Type I Errors)
Test
VIN
Vehicle
Type
Model
Year
HC
(g/mi)
IM240
Standard
IM147
Standard
CO
(g/mi)
IM240
Standard
IM147
Standard
NOx
(g/mi)
IM240
Standard
IM147
Standard
I Ml 47, test 1
3G4AG54N5PS614754
LDGV
90-93
0.72
0.8
7.27
15
3.84
2.2
IM147, test 2
3G4AG54N5PS614754
LDGV
90-93
0.04
0.8
0.28
15
2.7
2.2
IM147, test 3
3G4AG54N5PS614754
LDGV
90-93
0.01
0.8
0.56
15
1.8
2.2
IM240 test
3G4AG54N5PS614754
LDGV
90-93
0.01
0.8
0.29
15
2.11
2
IM147, test 1
JT4RN50R4G0121113
LDGT1
86-89
1.55
1.7
20.32
43
1.08
2.9
IM147, test 2
JT4RN50R4G0121113
LDGT1
86-89
1.27
1.7
21.03
43
0.81
2.9
IM147, test 3
JT4RN50R4G0121113
LDGT1
86-89
1.39
1.7
25.86
43
1.12
2.9
IM240 test
JT4RN50R4G0121113
LDGT1
86-89
1.68
1.6
26.86
40
1.25
4.5
IM240 Pass - Third IM147 Test Fail (Tvoe II Errors)
Test
VIN
Vehicle
Type
Model
Year
HC
(g/mi)
IM240
Standard
IM147
Standard
CO
(g/mi)
IM240
Standard
IM147
Standard
NOx
(g/mi)
IM240
Standard
IM147
Standard
IM147, test 1
1MEBP95F9EZ649996
LDGV
81-85
0.88
0.8
19.25
16
2.26
2.3
IM147, test 2
1MEBP95F9EZ649996
LDGV
81-85
0.43
0.8
12.57
16
1.76
2.3
IM147, test 3
1MEBP95F9EZ649996
LDGV
81-85
0.42
0.8
16.01
16
1:72
2.3
IM240 test
1MEBP95F9EZ649996
LDGV
81-85
0.38
0.8
12.77
15
1.43
2
IM147, test 1
1G1AW81R9H6121416
LDGV
86-89
0.21
0.8
3.7
16
2.65
2.2
IM147, test 2
1G1AW81R9H6121416
LDGV
86-89
0.41
0.8
12.66
16
3.04
2.2
IM147, test 3
1G1AW81R9H6121416
LDGV
86-89
0.07
0.8
2.42
16
2.22
2.2
IM240 test
1G1AW81R9H6121416
LDGV
86-89
0.04
0.8
0.94
15
1.77
2
IM147, test 1
1B3XC56R8LD901740
LDGV
90-93 .
0.37
0.8
4.72
15
2.71
2.2
IM147, test 2
1B3XC56R8LD901740
LDGV
90-93
0.16
0.8
4.07
15
2.38
2.2
IM147, test 3
1B3XC56R8LD901740
LDGV
90-93
0.16
0.8
5.14
15
2.45
2.2
IM240 test
1B3XC56R8LD901740
LDGV
90-93
0.15
0.8
4.28
15
1.95
2
-------
• The emissions for the second vehicle were within 0.02 g/mi of the IM147
NOx cutpoint, but the emissions of the vehicle were fluctuating. The
emissions first increased from the first to the second test, and then decreased
again. As can be seen from the HC and especially the CO emissions rates,
there was significant variability in all pollutants.
• The last vehicle in this group was over the IM147 NOx cutpoint by more than
10% and passed the IM240 cutpoint by 2.5%. This vehicle also had
fluctuations in emissions, decreasing from the first to the second IM147 test,
and then increasing between the second and third tests.
The impact these five vehicles had on the differences in failure rates between the IM147
and IM240 can be seen in Figures 4-1 and 4-2. The figures show the failure rates as a
function of vehicle type and model year groupings for all three pollutants, and overall* for
the three IM147 tests and the IM240 using final standards. In Figure 4-1, the overall
failure rates for all pollutants by the second IM147 test are very close the IM240. It is not
expected that the failure rates will be the same on the first IM147 test as on the IM240
because the vehicle may still be warming up; however, the agreement should improve as
the vehicle warms up, as observed in the results. LDGV failure rates for all three
pollutants decreased from the first IM147 test to the third, as expected. LDT2s predicted
well for CO from the first test, overpredicted for the first test for HC and NOx, but agreed
well with IM240 failure rates for the third IM147 test.
Looking at the failure rates by pollutant and model year (Figure 4-2), a similar trend is
seen among the pollutants, in that the first IM147 test failure rate is about 30% to 40%
higher than the IM240, but agrees well after the second IM147 test. For 1981-1985
model year vehicles, the agreement for HC is good by the second IM147; however, the
IM147 has lower failure rates for CO, and the agreement for NOx is not good until the
third IM147 test. For 1986-1989 model year vehicles, there was a drop from the first to
second test, but no change between the second and third tests. The failing sample for
1990-1993 model year vehicles was small, but the agreement was good after the third test
for all three pollutants, with the largest variation in NOx emissions for the first IM147
test. There were too few failing 1994 and newer vehicles to determine if the cutpoints
agree well.
The impact of changes in test times can also be seen in Table 4-5 by the cutpoints that are
applied. The sample-weighted average test time for the three groups increased with the
stringency of the standards. The overall average test times for the three groups were 172
seconds per test for the startup cutpoints, 185 seconds for the intermediate cutpoints, and
206 for the final cutpoints. This is a 20% increase in test time going from startup
cutpoints to final cutpoints. Note, however, that those test times do not include the
*
Because of the small sample size, some of the failures are represented by only a few vehicles, especially
for the LDGT2 vehicle type. It is important to note that the failure rates shown for the "all" categories in
both figures were computed by simple averaging of the test results from the 300 vehicles contained in the
sample. Thus, they reflect the distribution of vehicle model years and vehicle types included in the sample,
and have not been adjusted for the fleet characteristics in Arizona.
-40-
-------
Figure 4-1
Failure Rates for Three Consecutive IM147 Tests Followed by an IM240 Test
by Vehicle Type Using Final Cutpoints
CO Fail Rates During Test Sequence
40% .
—30%
K
= 20%
10%
0%
:jd
LDTl LDT2
Vehicle Type
glstlM 147
02nd IM 147
03«d IM 147
QFmallM 240
i
=
n ¦!
¦miiiiiii IS
NOx Fail Rates During Test Sequence
40%
30%
LOT 1 LDT2
Vehicle Type
gist IM 147
Q2nd IM 147
Q3rd IM 147
~ Final IM 240
Overall Fail Rates During Test Sequence
n
gist IM 147
PC LOT 1 L0T2 All
Vehicle Type
-------
Figure 4-2
Failure Rates for Three Consecutive IM147 Tests Followed by an IM240 Test
by Model Year Group Using Final Cutpoints
I
to
HC Fail Rates During Test Sequence
86-89 90-93
Model Year Group
3994+
CO Fail Rates During Test Sequence
m 1st IM 147
~ 2nd IM 147
~ 3rd IM 147
QFinal IM 240
86-89 90-93
Mode I Year Group
100%
80%
2 60%
«
NOx Fail Rates During Test Sequence
X
O
z
40% . _
20%
0%
81-85
86-89 90-93
Model Year Group
fllStlM 147
~ 2nd IM 147
~ 3rd IM 147
~ FinallM 240
1
EE
n
¦ -
l=U, ¦
Overall Fail Rates During Test Sequence
100% t I :
81-85 86-89 90-93 1994+
ModelYear Group
-------
impacts of fast-pass methodologies, nor do they incorporate the impact of a retest
algorithm to identify vehicles inadequately preconditioned.
As expected, test times by model year group were higher for older model year vehicles,
since these vehicles have higher emissions and are more likely to fail the test, requiring
another test. For startup and intermediate cutpoints, the test time for newer vehicles was
40% lower than for older vehicles; for final cutpoints, newer vehicles' test times were
over than 50% lower than test times for older vehicles.
Modal Predictive Fast Pass Analysis
The first technique applied to the new IM147 test to help shorten test time was to apply
fast pass standards to allow those vehicles with emissions well below the standards to
complete the test early. Currently, every I/M program performing IM240 testing is using
a fast-pass algorithm to reduce average test time. In previous work for EPA,7 Sierra
evaluated methodologies used to develop fast pass standards, adopted the most reasonable
approach, and modified it slightly to improve its performance. The methodology for
development of fast-pass standards is reviewed briefly here because it is important for
understanding its application. In that approach, full-duration test scores are regressed
against emissions during particular segments, or modes, of the IM147. Thus, at the
conclusion of each mode of the test, the full-duration score can be estimated by applying
the coefficients developed from the regression analysis. In all, 18 regression models were
developed in this effort: six each for HC, CO, and NOx, with these six representing
composite and Phase 2 IM147 standards for LDGVs, LDGTls, and LDGT2s.
Although current methodologies used to develop fast-pass cutpoints provide workable
solutions to establishing fast-pass cutpoints, they do not make full use of the information
collected on a second-by-second basis as the test is occurring. Rather, the fast-pass
decision is made simply by comparing cumulative emissions at each second to a
particular emission standard at each second. With the intent of improving the
performance of the current fast-pass methodology and standards (i.e., shorter test time
and fewer false passes), alternative approaches from Resources for the Future8 and the
New York Department of Environmental Conservation9 were evaluated. In reviewing
these approaches, the method that had the most appeal (both from an engineering and a
statistical perspective) was the modal regression technique developed by NYDEC. Thus,
the methodology developed by NYDEC formed the basis of the approach adopted by
Sierra and this approach was used to generate fast-pass IM147 standards in this study.
The methodology divides the test cycle into modes that are cumulatively evaluated as the
test progresses. For the IM147, the last 13 modes that were developed for the IM240 in
the previous study were used. The last four of these modes were used for the Phase 2
portion. These modes are illustrated in Figure 4-3.
-43-
-------
Figure 4-3
IM147 Test Modes Used for Fast-Pass Cutpoint Development
Time (sec)
The modal regression technique developed in the previous study was then applied to the
300 vehicles in the current data set to develop a set of fast-pass regression models for the
IM147 test. Consistent with the methodology presented above, regression coefficients
were generated separately for LDGVs and LDGTl&2s with model year groups to match
the previous analyses.
As mentioned earlier, regression coefficients were developed for HC, CO, and NOx for
both the composite IM147 and for Phase 2. The first mode at which a pass/fail decision
was allowed was mode 4 (which ends at second 33 of the IM147) for a composite IM147
prediction, or mode 8 (which ends at second 67 of the IM147) for a Phase 2 prediction.
The regression coefficients for HC, CO, and NOx are given in Appendix A for both the
full IM147 and Phase 2 for all vehicle classes, model year groups, and pollutants.
The vehicles from this study were used to calculate excess emissions identified and
change in test time both with and without the fast-pass regression models. This analysis
was performed using the startup, intermediate, and final IM147 and IM240 HC, CO, and
NOx standards. The results are shown in Table 4-7 for comparison. Test time
comparison for IM147 tests was based on three tests in a row or until passing; IM240
was based on two tests in a row or until passing.
Excess emissions were calculated in the same manner as for development of new
standards. Excess emissions identified were the emissions on the IM240 for failing
vehicles above the standard and also failing the IM147.
-44-
-------
Table 4-7
Comparison of Fast-Pass Effectiveness for the IM147
Impact on Test Time and Excess Emissions Lost
Vehicle
Class
Model
Year
Group
IM147
Test Time
without
Fast Pass
(seconds)
IM240
Test Time
with
Fast Pass
(seconds)
IM147 with Fast Pass
Number
in Test
Sample
Mean Test
Time
(seconds)
% Excess Emissions
Identified"
HC
CO
NOx
Startup Cutpoints
LDGV
81 - 84
294
221
13
249
100
100
100
85-89
180
114
45
86
100
100
-
1990+
159
50
133
56
100
100
100
LDGT
1&2
81 - 84
239
138
8
146
93.8
-
100
85-89
185
111
. 27
62
0.0
0.0
-
1990+
159
53
74
56
-
-
100
Weighted average
172
75
72
93.8
95.1
100
Intermediate Cutpoints
LDGV
81-84
328
270
13
289
96.9
91.4
98.1
85 - 89
209
160
45
136
100
100
98.1
1990+
166
71
133
64
95.3
100
100
LDGT
1&2
81-84
294
219
8
221
100
100
55.5
85-89
201
160
27
109
90.3
100
0.0
1990+
163
75
74
59
0
-
100
Weighted average
185
106
92
98.0
97.8
89.5
Final Cutpoints
LDGV
81 - 84
362
308
13
348
100
100
100
85 - 89
255
218
45
204
98.3
100
100
1990+
171
77
133
78
100
100
96.4
LDGT
1&2
81-84
368
255
8
343
100
100
100
85-89
289
195
27
159
97.2
77.3
100
1990+
165
79
74
72
100
-
100
Weighted average
206
122
121
99.2
94.8
99.6
a indicates there were no failures in this group for the IM240 and therefore no excess emissions for
the IM147 to identify.
-45-
-------
The non fast-pass test times are from the analysis of the new standards presented in the
previous portion of this section (the fourth column in Table 4-5). Averaging these by the
number of vehicles in each class in each model year group of the fleet, the overall effect
on the total test time can be determined. Test times for the IM240 with fast-pass come
from a previous analysis of 26,000 vehicles that were tested in Arizona as part of the "2%
Random Sample."10 Because the 2% random sample includes only initial tests, the test
times had to be adjusted to include retests; this adjustment was based on analysis of
Arizona test data.11 On average, vehicles in the Arizona test data that failed the first
IM240 test and were subjected to another test went 110 seconds on the retest, so this test
time was added to the 240 seconds for all failing vehicles. In addition, the average
IM240 test times by model year group were then weighted by the fleet distributions in the
sample vehicle fleet from this study to allow for an even comparison of test times.
As shown in Table 4-7, substantial test time reductions occur as a result of implementing
fast-pass algorithms. Reductions of 58% for startup cutpoints, 50% for intermediate
cutpoints, and 41% for final cutpoints were observed. Comparison of these test times to
test times for the use of fast-pass for IM240 testing previously reported to EPA shows the
IM147 has lower test times for startup and intermediate cutpoints (72 seconds for the
IM147 versus 75 seconds for the IM240 for startup cutpoints, and 92 seconds for the
IM147 versus 106 seconds for the IM240 for intermediate cutpoints). Test times for final
cutpoints were almost the same (121 seconds for the IM147 versus 122 seconds for the
IM240).
Overall excess emissions identified by the IM147 using the above fast-pass methodology
were all above 95% with only one exception (for intermediate NOx cutpoints). At the
same time, there were excess emissions for individual vehicle model years that the IM147
did not identify well. The majority of the loss in excess emissions was for LDGTs in the
1985 to 1989 range. For startup standards, the IM147 lost 100% of the excess emissions
for this model year/vehicle type grouping. This was due to the small sample size used in
this analysis. There was only one vehicle that failed the IM240; however, that vehicle
falsely passed the IM147, and therefore there was a loss in excess emissions of 100%.
However, the excess emissions lost for this vehicle were less than 5% of the overall
excess HC and CO emissions of the fleet. If a larger sample were tested, it is expected
that these types of anomalies would be eliminated.
Modal Predictive Retest Analysis
The second method evaluated to reduce test time was the development of algorithms to
predict if a vehicle would benefit from a retest. If a vehicle were not adequately
preconditioned (warmed up) before the test (e.g., due to being in a long queue), then the
vehicle could falsely fail the emissions test. Although the "two-ways-to-pass" standards,
which use the first portion of the IM test as preconditioning for the second half of the test,
work for some vehicles, previous studies conducted by Sierra revealed that inadequate
preconditioning can be responsible for up to 25% false failures.12,13 As shown in those
earlier studies, modal data from various sections of the test can be used to predict with at
least 80% accuracy if a vehicle that fails would benefit from a retest due inadequate
preconditioning. If a vehicle would continue to fail repeated tests (i.e., it did not fail due
-46-
-------
to a lack of preconditioning), then retesting the vehicle is an inefficient use of testing
time. If failing vehicles that will continue to fail can be discriminated from those
vehicles that would benefit from another test (i.e., vehicles that failed due to a lack of
preconditioning but would pass an additional test), retesting of the fail/fail vehicles could
be avoided and average test time could be reduced.
Using a similar approach as was used in the previous studies,14 the composite emission
rate, the Phase 2 emission rate, and the concentrations during certain sections of the test
were evaluated to determine if relationships exist in the data that can help predict if a
vehicle would benefit from a retest. To develop the algorithms for the IM147, the
sections of the test that were initially evaluated based on both mass and concentration
included seconds 43 to 48, 54 to 59, 82 to 106, and 116 to 121 (assuming the first second
of the test is 0), as shown in Figure 4-4. Proposed final cutpoints were used to determine
pass-fail status. Due to the limited sample size, LDGTls and LDGT2s were combined
for this analysis and failures between either IM147 test one and two, or IM147 test two
and three, were treated the same for developing the retest criteria. Based on these
conditions, data from 111 failed tests (75 LDGVs and 36 LDGTs) were used to develop
the retest algorithms.
Figure 4-4
Modes Used For Development Of Retest Algorithms
The logic applied to LDGVs and LDGTs to determine if a vehicle should be retested is
shown in the flow diagrams presented in Figures 4-5 and 4-6. The first criterion applied
was that if a vehicle failing for all three pollutants would not be retested. This condition
occurred in 6.7% of the failing LDGVs and 5.7% of the LDGTs.
-47-
-------
Figure 4-5
l/M 147 Retest Predictive Model Logic
LDGV
-48-
-------
Figure 4-6
l/M 147 Retest Predictive Model Logic
LDGT1 & 2
-49-
-------
The next criterion was to evaluate emissions from the whole test or Phase 2 of the test,
relative to the emissions standards. If the vehicle was within a certain percent of the
standard, a retest was usually warranted. This worked well as an initial criterion for all
vehicle classes and pollutants with the exception of HC for LDGVs. For LDGTs, this
was the only criterion needed; all others had one additional criterion. The remaining HC
and CO criteria were based on either Phase 2 or Phase 2 divided by the composite
emissions compared to their standards. For NOx, the additional criterion was based on a
percent concentration reduction between mode A and mode C (the ratio of the average
concentration during mode 3 shown in Figure 4-4 and mode 1 shown in Figure 4-4),
which was an indication of warm-up occurring as this percent decreased.
The final logic that was applied was that if a vehicle failed for more than one pollutant, a
retest for every pollutant had to be indicated. Therefore, even if the vehicle appeared to
be warming up, if it was not going to be enough to pass the standards for each pollutant,
the vehicle was not retested.
The accuracy of these algorithms in predicting the need for a retest correctly is shown in
Table 4-8.
Table 4-8
Evaluation of Retest Criteria (Based on Final Cutpoints)
Sample
size
% correct
Retested when
vehicle would
still fail
Did not retest
when vehicle
would pass
LDGV
Fail HC, CO, & NOx
5
100 %
0.0 %
0.0 %
HC failures
48
95.8 %
2.1 %
2.1 %
CO failures
25
92.0 %
4.0 %
4.0 %
NOx failures
48
81.3%
10.4%
8.3 %
All criteria for all failures
75
86.6 %
6.7 %
6.7 %
LDGT
Fail HC, CO, & NOx
2
100%
0.0 %
0.0 %
HC failures
18
94.4 %
5.6%
0.0 %
CO failures
9
100%
0.0 %
0.0 %
NOx failures
24
95.8 %
4.2 %
0.0 %
All criteria for all failures
36
97%
2.8 %
0.0 %
LDGV and LDGT (Sample Fleet Weighting)
All criteria for all failures
111
90.0%
5.4 %
4.5 %
-50-
-------
As shown in Table 4-8, applying the combined criteria to the failing vehicles in the
sample fleet resulting in a overall correct prediction rate (of when a vehicle needed a
retest) of 90%. The accuracy of the algorithms were, however, better for LDGTs than for
LDGVs. This may be an artifact of having only a limited sample set to use to develop
and evaluate the algorithms. Because there were fewer data for LDGTs, it was easier to
fit a model with low or no false failures or false passes. As more test data become
available, the criteria should be re-evaluated and revised.
The results on test time and excess emissions identified by applying the retest criteria to
the full sample set are shown in Table 4-9. As shown in the table, application of the
retest algorithms resulted in significant reductions in test time from the reductions
produced by using fast-pass, including 24% for startup cutpoints, 28% for intermediate
cutpoints, and 33% for final cutpoints. The trend of higher reductions in test times for the
startup cutpoints that occurred with the application of fast-pass standards was reversed
with the application of the retest algorithms. This caused the overall test time reductions
for the three sets of cutpoints to come closer together. The reduction in test time from not
using either fast-pass or retest algorithms to using both was 68% for startup cutpoints,
64% for intermediate cutpoints and 61% for final cutpoints.
As can be seen from Table 4-9, implementing the retest algorithm identified more excess
emissions than did use of fast pass cutpoints alone. A retest algorithm cannot cause a loss
in excess emissions identified because it does not make decisions about passing a vehicle,
it is used only to determine if a failing vehicle should receive a retest. In fact, there were
several cases with the current data set where the retest logic prevented a vehicle from
undergoing a second test in which the vehicle would have been falsely fast-passed. This
actually prevented excess emissions from being lost. An example is the intermediate CO
cutpoints with and without the retest algorithm. With only fast pass procedures
implemented, the excess emission identified for CO was 97.8%, but with the both fast
pass and retest algorithms, the excess emissions identified returned to 100%.
To more objectively evaluate the accuracy of the retest predictive algorithms, they were
applied to another sample of 101 vehicles tested on the IM147* in a previous study.15
This evaluation included an assessment of false-pass or false-fail occurrences, as well as
excess emissions lost and average test times. In this analysis, the fast-pass regressions
were also used because it is assumed that both of these techniques will be used in real
operations. The results of the evaluation are shown in Table 4-10.
*
The incorporation of the retest algorithm will tend to minimize the impact of the 30-minute idle period
prior to the testing of each of these vehicles. As a result, these data can be analyzed with less concern that
the results (in terms of % excess emissions identified) will be significantly biased relative to the actual
Arizona test fleet.
-------
Table 4-9
Comparison of Retest Predictive Algorithm Effectiveness for the IM147
Impact on Test Time and Excess Emissions Lost
Vehicle
Class
Model
Year
Group
Test Time
Without
Retest
Algorithm
(seconds)
IM147 with Retest Algorithm
Number in
Test
Sample
Mean Test
Time (seconds)
% Excess Emissions Identified3
HC
CO
NOx
Startup Cutpoints
LDGV
81 - 84
249
13
105
100
100
100
85-89
86
45
63
100
100
-
1990+
56
133
47
100
100
100
LDGT
1&2
81 - 84
146
8
109
100
-
100
85-89
62
27
54
93.8
0.0
-
1990+
56
74
48
-
-
100
Weighted average
72
55
97.2
95.1
100
Intermediate Cutpoints
LDGV
81 - 84
289
13
125.15
100
100
98.1
85 - 89
136
45
92.24
100
100
98.1
1990+
64
133
53
95.3
100
100
LDGT
1&2
81 - 84
221
8
137
100
100
76.1
85-89
109
27
79
90.3
100
0.0
1990+
59
74
52
0
100
Weighted average
92
66
98.4
100
93.2
Final Cutpoints
LDGV
81-84
348
13
167
100
100
100
85-89
204
45
118
98.3
100
100
1990+
78
133
61
100
100
96.4
LDGT
1&2
81-84
343
8
123
100
100
100
85 - 89
159
27
118
97.2
89.2
100
1990+
72
74
60
100
-
100
Weighted average
121
81
99.2
97.5
99.6
a indicates there were no failures in this group for the IM240 and therefore no excess emissions for
the IM147 to identify.
-52-
-------
Table 4-10
Evaluation of Retest Criteria with 101 Vehicle Sample
Based on Final Cutpoints and Exempting 1992+ Vehicles
Vehicle class
Number
in Test
Sample
%
correct
Retested when
vehicle would
still fail
Did not
retest when
vehicle
would pass
% Excess Emissions
Identified
HC
CO
NOx
LDGV
31
80.6 %
3.2 %
16.1 %
94.1
98.6
100
LDGT1&2
14
64.2 %
14.3 %
21.4%
30.6
30.4
71.1
Weighted average of
LDGV and LDGT1&2
45
75.6 %
6.6 %
17.8 %
74.2
89.8
92.1
The results show that the number of errors for LDGTl&2s were larger than for LDGVs.
This is most likely because the LDGT1&2 retest criteria are based on fewer data than the
LDGV retest criteria, so they do not work as well. Comparing the weighted average
number of vehicles that were retested when the vehicle failed versus those that were not
retested but that would have benefitted from a retest shows the criteria appear to be biased
toward not testing vehicles that could benefit from another test. The impact of this may
be to cause vehicle owners to obtain unnecessary vehicle repairs. These criteria for retest
will need to be examined further when more data are available to further refine of the
retest models.
The results of the analysis of the percent of excess emissions identified shows the largest
emissions losses were for HC, then CO. All NOx emissions were identified in the small
sample. The losses in excess HC emissions identified are much higher than in the 300
vehicle sample. The excess NOx emissions loss is less, and the CO loss is in the same
range as for the other sample. Again, the variation is likely the result of the small sample
size. Further improvement of the retest algOTithms will require more data.
Model Year Exemptions
A very efficient method to reduce the overall test time is to simply remove from testing a
portion of the fleet that would likely pass the emissions test anyway. Currently, EPA has
published draft guidance16 on the use of "Clean Screening" methods for identifying
potentially clean vehicles to be exempted from testing. The methods discussed in the
report and currently under evaluation including the use of model year exemptions, a low
emitter profile, and remote sensing to identify likely clean vehicles. There has been much
debate over the relative merits of these techniques, and the process is continuing.
*
See the U.S. EPA web site section on Clean Screening, http://www.epa.gov/orcdizux/rsd.htm for more
details on the debate.
-53-
-------
Of the proposed methods, a model year exemption is the easiest to implement and should
have virtually no direct cost to implement. In addition, new vehicles are historically low
emitters, in part because of emissions control system warranty requirements imposed on
new vehicles and in-use compliance programs that are highly effective in enforcing
emissions system durability requirements. A low emitter profile is an extension of a
model year exemption, in which other characteristics of vehicles that are predictive of
low emissions (e.g., make, model, engine family, etc.) are also used to determine which
vehicles should be exempted from testing. This method can exempt more vehicles than a
model year exemption alone (new vehicles plus older vehicles which the model believes
are clean), but it requires a database of vehicle emissions tests in order to develop the
profiles (which has some cost). The use of remote sensing to identify low emitting
vehicles has both significant technical accuracy problems and program costs. Part of the
reason for developing methods to reduce test time is to improve the efficiency and cost
effectiveness of emissions testing. For this reason, a model year exemption seems to be a
very reasonable method to use, and it was considered as part of the present study.
This issue was first analyzed by removing the excess emissions in the IM240 from the
excess emissions identified by the IM147, for the model years being considered for
exemption. All results are compared to a baseline that includes the fast-pass and retest
algorithms. Vehicles in the present study ranged from 1981 to 1998 model years.
However, because only one 1998 model year vehicle was included in the data set, it was
assumed that a "current plus five" model year exemption would include model years
1997 through 1992. The results for exempting these model years (with the use of fast-
pass and retest algorithms) is shown in Table 4-11.
When exempted vehicles are included in the analysis, the reductions in mean test time
from the use of all three techniques are 82% (from 172 seconds per test to 31 seconds per
test) for startup outpoints, 78% (from 185 seconds per test to 41 seconds per test) for
intermediate cutpoints, and 73% (from 206 seconds per test to 55 seconds per test) for
final cutpoints. When only those vehicles arriving at the test lanes are considered (i.e.,
the exempted vehicles are not included with a "test time" of zero), the actual mean test
times are 65 seconds for startup cutpoints, 86 seconds for intermediate cutpoints, and 115
seconds for final cutpoints. However, these results are for the sample fleet. The model
year distribution of this fleet may not match the model year distribution of all fleets. It is
important to note that the estimated change in test time will vary with different fleets due
to differences in the vehicle age distribution.
For the test sample under the startup cutpoints, the excess emissions lost was greatest for
CO; under the intermediate cutpoints, it was greatest for NOx. For final cutpoints, the
maximum excess emissions lost (7.9%) was for CO.
-54-
-------
Table 4-11
Comparison of Exempting 1992+ Model Years (Current + 5)
Impact on Test Time and Excess Emissions Lost
Vehicle
Class
Model
Year
Group
Test Time
Without
Model Year
Exemptions
(seconds)
IM147 with Model Year Exemptions
Number
in Test
Sample
Mean Test Time (seconds)
% Excess Emissions
Identified"
All
Vehiclesb
Non-Exempt
Vehicles
HC
CO
NOx
Startup Cutpoints
LDGV
81 - 84
105
13
105
105
100
100
100
85 - 89
63
45
63
63
100
100
-
1990+"
47
133/37°
-
57
89.8
36.3
100
LDGT
1&2
81 - 84
109
8
109
109
93.8
-
100
85 - 89
56
27
54
54
100
0.0
-
1990+"
48
74/14c
-
55
-
-
67.9
Weighted average1*
55
31
65
95.2
81.7
88.5
Intermediate Cutpoints
LDGV
81 - 84
125
13
125
125
100
100
98.1
85-89
92
45
92
92
100
100
98.1
1990+"
53
133/37c
-
68
79.6
39.4
100
LDGT
1&2
81 - 84
137
8
137
137
100
100
76.1
85 - 89
79
27
79
79
90.3
100
0.0
1990+b
52
74/14c
-
59
0.0
-
61.9
Weighted averaged
66
41
86
96.0
91.5
84.1
Final Cutpoints
LDGV
81 - 84
167
13
167
167
100
100
100
85-89
118
45
118
118
98.3.
100
100
1990+"
61
133/37c
-
100
73.3
43.2
96.4
LDGT
1&2
81 - 84
123
8
123
123
100
100
100
85 - 89
118
27
118
118
97.2
89.2
100
1990+"
60
74/14c
-
87
100
-
58.3
Weighted averaged
81
55
115'
95.9
92.1
92.4
a indicates there were no failures in this group for the IM240 and therefore no excess emissions for the
IM147 to identify.
b Test time shown for 1990+ vehicles is for those that were tested and not exempted.
c Values represent the number of 1990+ vehicles in the sample / the number which were not excluded by the
model year exemption.
d Weighted average for all vehicles includes a test time of 0 for vehicles exempted; weighted average for non-
exempt vehicles does not include the newer models that have been exempted in the calculation of test time.
-55-
-------
Table 4-12 summarizes the excess emissions identified by model years exempted for the
final IM147 cutpoints. The excess emissions lost for all three pollutants are the same if
1994+ or 1993+ vehicles are exempted. This shows the effect of the small sample size -
in a larger sample, some change would be expected between model years exempted.
Table 4-12
Excess Emissions Identified versus Model Years Exempted3
HC
CO
NOx
1994 +
99.2 %
97.5 %
92.4 %
1993 +
99.2%
97.5 %
92.4%
1992 +
95.9 %
92.1 %
92.4 %
1991 +
86.8 %
87.9 %
87.9. %
a Based on final IM147 cutpoints.
In the present case, there were only a few vehicles of each model year, and it just happens
that there were no 1993 vehicles with excess emissions.
Model Year Exemptions Over All Model Years
Since the I/M test data used in this project are from Arizona where IM240 testing is not
conducted on 1980 and later model year vehicles, the percent of excess emissions lost
from exempting newer model years as shown above, is based only on the 1981+ model
years. A more representative comparison would be to consider these emissions losses
relative to the total excess emissions identified for all vehicles in the I/M program,
including pre-1981 model years. To evaluate the emissions impact based on the excess
emissions of an entire fleet, data from the Wisconsin I/M program (where IM240 testing
for 1968 and newer model year vehicles has been in effect since December 1995) were
used. Because no IM147 data have been collected in Wisconsin, this analysis was based
on the IM240 data. However, this analysis should serve as an indicator of the excess
emissions lost when the entire fleet is considered.
In the Wisconsin IM240 program, fast-pass algorithms are used, and therefore a random
sample of full duration IM240 tests is not available. Because of this, the partial test
scores had to be adjusted to composite (full test) emissions. To calculate the composite
emissions, cumulative 30-second HC, CO, and NOx emissions were tabulated for all of
the two-percent random sample tests conducted in Arizona from April through June
*
A random sample of 1981+ LDVs are subjected to full-duration IM240 tests in the Arizona IM240
program.
-56-
-------
1996, to be used in developing of the correlation equations. The April through June time
period was selected because of the typically moderate range of temperatures seen in both
Arizona and Wisconsin during that portion of the year. Next, linear regression equations
defining the relationship between cumulative 30-second and composite gram-per-mile
emissions were developed. Three sets of equations, which characterize the model year
groupings used in the rest of the analysis, were developed using this methodology.
The regression equations developed from the Arizona data set were applied to the
30-second scores from the Wisconsin IM240 Program data* collected during the same
April-June 1996 time frame. Composite (full duration) IM240 scores were extrapolated
from the Wisconsin cumulative 30-second emissions using the equations described
above. Excess emissions for each test were then determined by applying the appropriate
model-year EPA final cutpoints to the composite IM240 estimates. The excess emissions
were then averaged by model year, and weighted using MOBILE5a travel fractions.
Figure 4-7 graphically displays the results of the analysis. As shown in the figure,
exempting the latest five model years (1991-1995) has a relative insignificant effect (i.e.,
approximately 2% or less of each pollutant) on the cumulative excess HC, CO, and NOx
emissions identified when all model years are considered. The figure also shows that
these five model years account for roughly 45% of the total 1968 and later LDV
population subject to the program.
The fraction of excess emissions identified begins to decline more rapidly with the
exemption of the 1990 model year. However, up to nine model years (1987-1995) can be
exempted before the decline steepens appreciably for all three pollutants with decreasing
model year. This effect on cumulative excess emissions identified is also shown
tabularly in Table 4-13 for the latest 10 model years. The data presented in both the table
and the figure provide an approximation of the relative effect on excess emissions
identification due to exempting various ranges of newer model year vehicles from IM240
testing, when all model years are considered. Effects of a similar magnitude would be
expected in a program utilizing the IM147 test procedure.
Summary
The IM147 can produce significant reductions in test times compared to the IM240,
especially with the use of fast-pass procedures, retest algorithms, and new model year
vehicle exemptions. The overall losses in emissions are small and therefore the IM147
can produce cost-effective reductions in test time.
The Wisconsin IM240 sample included roughly 127,000 vehicles.
-57-
-------
Figure 4-7
Cumulative Excess IM240 Emissions Identified by Model Year,
Based on EPA's Final Cutpoints (1996 Wisconsin Data)
100%
T3
0)
£
'<•5
C
0)
2
M
c
o
O
X
Ui
80%
~ Population
60% -
40%
20%
0%
95 94 93 92 91 90 89 88 87 86 85 84 83 82 81 80 79 78 77 76 75 74 73 72 71 70 69 68
M odelYear
-------
Table 4-13
Effect of LDV Model Year Exemptions on Excess Emissions Identification
Cumulative % of Excess Emissions % of Vehicle
Model Year Population
HC CO NOx
1995
0.1%
0.0%
0.1%
8.3%
1994
0.3%
0.1%
0.2%
17.1%
1993
0.5%
0.3%
0.2%
25.7%
1992
1.2%
0.8%
0.4%
34.5%
1991
2.1%
1.4%
0.5%
43.1%
1990
5.2%
3.3%
2.6%
51.2%
1989
8.9%
5.6%
5.3%
57.1%
1988
13.1%
9.2%
7.8%
62.3%
1987
18.3%
14.2%
10.5%
67.7%
1986
28.9%
26.8%
20.9%
72.9%
Test Times - The changes in test time as each of the methods to reduce test time were
applied can be seen in Table 4-14. As the table shows, up to a 68% reduction in test time
can still be achieved under the final standards, when using all three test time reduction
techniques (i.e., fast-pass, retest algorithm, and model year exemptions). The largest
reduction was achieved by implementing fast-pass procedures. This is because the
majority of vehicles pass, and many are very clean and can pass out of the emissions test
quickly.
Excess Emissions Identified - Excess emissions identified for each pollutant are shown in
Table 4-15. There are some cases where excess emissions were "lost" by using one
method to reduce test time, and then "gained" back by the next technique that was
applied. This was in part because the techniques overlap in the way they work, i.e., the
fast-pass procedures could have resulted in a loss of excess emissions from a retest, but
the retest algorithm did not allow for the retest to occur.
Overall, total excess emissions lost by switching to the EM 147 are low, up to the point
where model years are exempted. Even if the current plus five model years are exempted,
the IM147 retains over 92% of the excess emissions identified by the IM240 under final
cutpoints.
-59-
-------
Table 4-14
Average IM147 Test Time (seconds)
Startup Cutpoints
Intermediate Cutpoints
Final Cutpoints
Cutpoint only, two possible retests"
172
185
206
Added fast-pass
. 72
92
121
Added retest algorithm
55
66
81
Added exemption
of 1992+
model years
All Vehicles
31
41
55
Non-Exempt
Vehicles
65
86
115
Overall %
reduction
in test time
All Vehicles
82%
78%
73%
Non-Exempt
Vehicles
62%
54%
44%
a If the vehicle fails the IM147, it is retested up to two more times to ensure the failure was not due to
a lack of preconditioning.
Table 4-15
Percent Excess Emissions Identified by the IM147 Test
Startup Cutpoints
Intermediate Cutpoints
Final Cutpoints
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
Cutpoint only, two
possible retests
100
100
100
99.2
100
93.6
99.6
98.2
99.6
Add fast-pass
96.5
95.1
100
98.8
97.8
89.5
99.6
95.5
99.6
Add retest algorithm
97.2
95.1
100
98.4
100
93.2
99.2
97.5
99.6
Add exemption of
1992+ model years
95.2
81.7
88.5
96.0
91.5
84.1
95.9
92.1
92.4
###
-60-
-------
5. DISAPPEARING VEHICLES
It is estimated that roughly 20% of vehicles that fail the Arizona IM240 test do not.
receive a passing test, resulting in a significant loss of potential SIP credit for the
program. In a previous study performed by Sierra for DEQ, it was discovered that a
significant fraction of vehicles failing the initial I/M test did not pass on a retest within a
four- to sixteen-month window following their initial test.17 The goal of this portion of
the work assignment was to evaluate the disposition of such disappearing vehicles,
estimate their excess emissions lost, and suggest solutions to address the problem. These
results will help to better define the nature and extent of the problem of retest
noncompliance in the Arizona program, which likely occurs in every other I/M program ,
in the U.S.
The first step in the analysis was to determine the fraction of those vehicles that continue
to be operated in the program area without receiving a passing I/M score. We had hoped
to receive information from the Arizona Motor Vehicle Department (through DEQ)
regarding whether those vehicles were being registered outside the area, or if their
registrations had lapsed. Unfortunately, the Motor Vehicle Department (MVD) was less
than cooperative in this request. As an alternative, remote sensing device (RSD) data
from Arizona's "Smog Dog" program were used to determine the fraction of non-
complying vehicles that continue to operate in the program area.
Definition of Data Groups
To perform the analyses that follow, Arizona I/M data collected from July to December
1997 were divided into three categories:
• Pass: Vehicles that passed the I/M test on the first attempt (initial test dates from
July 1, 1997, to September 30, 1997);
• Fail-Pass: Vehicles that failed the initial test (conducted between July 1, 1997,
and September 30, 1997), and passed within the next three to six months (i.e., by
December 31, 1997); and
• Fail-Fail: Vehicles that failed the initial test (conducted between July 1, 1997,
and September 30, 1997), and did not pass within three to six months of their
-61-
-------
initial test (i.e., by December 31, 1997). These vehicles are assumed to be driving
unregistered, and are considered the so-called "disappearing" vehicles.
This nomenclature is used through the rest of this discussion to describe the vehicles in
the study.
Comparison of I/M Data To RSD Observations
As outlined above, data regarding vehicles subjected to I/M testing in Maricopa County
from July through September of 1997 were used to identify vehicles that initially passed
or failed the I/M test. The initial test failures were tracked through the end of the year to
determine which of these vehicles had still not passed an emissions test (Fail-Fail
vehicles) within a three- to six-month time period following their initial test date (i.e., by
December 31, 1997). To determine whether the Fail-Fail vehicles were still being
operated in the area, the license plate numbers of these vehicles were compared to license
plate data of vehicles identified by remote sensors in Maricopa County from January 1
through March 31, 1998. The percentages of vehicles observed in the RSD data in Fail-
Fail and Fail-Pass categories are presented in Table 5-1 by model year groupings. Table
5-1 also presents the frequency that vehicles are seen on the road.
Table 5-1
Fraction of Vehicles Observed in the RSD Database
that Initially Failed an I/M Test in Maricopa County
Model Year
Group
Fail-Fail
July-Sept
1997
Fail-Fail
Observed
by RSD
Fail-Pass
July-Sept
1997
Fail-Pass
Observed
by RSD
Ratio of
Fail-Fail to
Fail-Pass
Pre-1975
823
2.7%
2,657
5.3%
51%
1975-1980
2,512
4.4%
8,690
7.0%
63%
1981-1984
2,148
6.2%
4,935
8.8%
70%
1985-1987
2,279
7.8%
6,662
10.2%
77%
1988-1991
1,154
9.2%
5,432
13.0%
71%
1992 +
282
13.8%
2,645
15.2%
91%
-------
A number of points can be made in reference to the data presented in Table 5-1:
• A smaller fraction of older vehicles were observed in the RSD data for both the
Fail-Fail and the Fail-Pass categories. This was expected because older vehicles
are driven less than newer vehicles, making them less likely to be identified by
RSD.
For all of the model year groups, the fraction of Fail-Fail vehicles observed by
RSD is less than that of the Fail-Pass vehicles. This implies that a certain fraction
of the Fail-Fail vehicles have been scrapped, parked, or otherwise removed from
the road. (Alternatively, the frequency of operation of these vehicles may have
been reduced.)
• The fraction of Fail-Fail vehicles not observed to be operating in the area
increases with vehicle age. Based on the results presented in Table 5-1, it appears
that about 50% of the Fail-Fail vehicles in the pre-1974 model year group have
been removed from the road. However, only 9% of the 1992 and later model year
Fail-Fail vehicles do not continue to operate on the road in Maricopa County.
• Although not presented in Table 5-1, the fraction of Pass vehicles (initial tests
from July 1997 to September 1997) observed in the January 1998 to March 1998
RSD data matched very closely the results for the Fail-Pass category. This would
appear to validate the use of the I/M data and the RSD data for this purpose.
An important element not considered in the analysis presented in Table 5-1 is waivered
vehicles. It is quite likely that a number of the Fail-Fail vehicles included in this analysis
did not receive a final passing score because they were granted a waiver. Data were
received late in the analysis regarding the waiver status of vehicles during the study
period so it could be determined if waivered vehicles accounted for a significant fraction
of the vehicles. Comparison of the data on waivered vehicles in Arizona to the
disappearing vehicles shows the waivered vehicles were not responsible. Waivered
vehicles accounted for only 0.71% of the disappearing vehicles.
An item of interest related to the analysis presented in Table 5-1 is whether the Fail-Fail
vehicles observed in the RSD database (and, therefore, operating in the area) continue to
make attempts to pass the I/M test. This was determined by merging the license plate
data from the Fail-Fail vehicles observed by RSD in the January 1998 to March 1998
time frame with the Arizona I/M database for the same time period. If vehicles appeared
in both databases, it was an indication that the vehicles were continuing with the I/M
process. The results of that analysis are summarized in Table 5-2, which indicates that
about 20% of the Fail-Fail vehicles operating in the area continue to be tested in an
attempt to receive a passing I/M score.
-63-
-------
Table 5-2
Fail-Fail Vehicles in the RSD Database that Continue
to Attempt to Pass an I/M Test
Model Year
Group
# of Fail-Fail
Observed
by RSD
Fail-Fail RSD Hits
I/M Tested
Jan to Mar 1998
Fraction of Fail-Fail
Continuing to be
I/M Tested
Pre-1975
22
4
18%
1975-1980
111
31
28%
1981-1984
134
18
13%
1985-1987
177
45
25%
1988-1991
106
27
25%
1992 +
39
3
8%
Comparison of Current Disappearing Vehicle Estimates to
Previous Estimates
As noted above, the issue of "disappearing" vehicles was first identified in a study
performed by Sierra for DEQ. In that study, which was based on an analysis of the 2%
Random Sample IM240 database, initial test results from January 1996 to December
1996 were merged with after-repair tests conducted from January 1996 through April
1997. (It was assumed that motorists intending to fully repair their vehicles would have
done so within a four-month time frame.) Based on an analysis of vehicles that received
at least one retest. the following non-passing after-repair percentages were obtained for
vehicles failing the initial test:
Fraction of Incomplete Repairs Based on DEQ Analysis
Model Year LDGV LDGT1 LDGT2
1981-1984 33% 20% 16%
1985-1989 23% 19% 15%
1990+ 9% 6% 7%
Although a slightly different methodology was used in the analysis presented in Table 5-1
(i.e., the entire database was analyzed rather than only the 2% Random Sample of IM240
tests, a minimum of one retest was not required, and a different time period was
Sierra chose to limit this analysis to vehicles that received at least one retest because of concerns over
whether the vehicles were coded correctly as being in the "Repair Sample" on their retest. If they received
at least one retest, then it is likely they were properly flagged on ensuing tests. However, this approach
may understate the magnitude of vehicles not receiving complete repairs in the Arizona program.
-64-
-------
analyzed), similar results were obtained. Using the model year groupings in Table 5-1,
the fraction of vehicles not passing their final test is summarized below.
Fraction of Incomplete Repairs Based on Current Analysis
Model Year All Vehicle Classes
Pre-1975
1975-1980
1981-1984
1985-1987
1988-1991
1992+
24%
22%
30%
25%
18%
10%
Although the model year groupings do not match exactly (a broader range of model years
was used in the DEQ analysis because of sample size considerations), similar results are
observed when comparing the two analyses. It is interesting to note that there is a peak in
incomplete repairs observed in the Table 5-1 data for the 1981-1984 model year group.
That is likely because those vehicles are subject to a more stringent I/M test than the older
vehicles (i.e., IM240 versus an idle/loaded test) and the waiver cost limit is greater (i.e.,
$100 forpre-1975; $300 for 1975-1979; and $450 for 1980 and later model year
vehicles), making it more difficult and more expensive to pass an I/M test. Given the
relative value of a 1981 model year vehicle, some motorists may choose to scrap or park
their vehicles (or continue to operate them unregistered), rather than perform the repairs
needed to pass an I/M test.
Conclusions
The analyses discussed above indicate that vehicles that continue to fail the I/M test after
a number of months (i.e., the Fail-Fail "disappearing" vehicles) are observed operating on
the road less frequently than their counterparts that received an initial I/M test in the same
time period. However, this is a function of vehicle age, with older vehicles being less
likely to continue to be operated (possibly scrapped, parked, or sold outside the area) than
newer vehicles. Using RSD data to infer operation frequency, it appears that about half
of the pre-1975 model year Fail-Fail vehicles do not remain in operation six to nine
months after their initial test, while nearly 90% of the 1992 and later model year Fail-Fail
vehicles remain on the road. Of those vehicles that do remain on the road, approximately
20% continue to attempt to pass the I/M test.
The fraction of initial test failures not receiving complete repairs (i.e., the Fail-Fail
vehicles) estimated in this analysis agreed very well with the results of Sierra's previous
analysis prepared for DEQ. Both analyses showed that older vehicles are more likely to
not be repaired completely, with about 30% of the 1981 to 1984 model year initial test
failures not receiving a passing score on the last I/M test.
Areas that may be investigated further to better define and develop recommendations to
reduce the number of Fail-Fail vehicles operating in the enhanced area include
(1) evaluating the potential effect of changes in the registration enforcement criteria in
-65-
-------
Arizona (currently they include a $300 fine with no "grace period"); (2) determining if
receiving a waiver is so difficult that motorists are driving unregistered instead of
attempting to receive a waiver; (3) determining if Fail-Fail vehicles are eventually being
registered out of the area but still driven in the area (there is anecdotal evidence of this);
and (4) continuing to work with DEQ to enlist the assistance of MVD to track down the
Fail-Fail vehicles. Unfortunately, Sierra's last correspondence with DEQ indicated that it
is unlikely that information from MVD will be forthcoming in the near future.
II11II
###
-66-
-------
6. REFERENCES
1. "Additional Study of Preconditioning Effects and Other IM240 Testing Issues,"
Prepared by Sierra Research for the U.S. Environmental Protection Agency,
Report No. SR98-02-01, February 2, 1998.
2. "Analysis of Alternate IM240 Cutpoints, Phase 2 Testing, and Exempting New
Vehicle Models on Test Duration and Projected I/M Benefits," Prepared by
Sierra Research for the Arizona Department of Environmental Quality, Report
No. SR98-05-01, May 12, 1998.
3. "Additional Study of Preconditioning Effects and Other IM240 Testing Issues,"
Prepared by Sierra Research for the U.S. Environmental Protection Agency,
Report No. SR98-02-01, February 2, 1998.
4. "Description and Documentation for Interim Vehicle Clean Screening Credit
Utility , Draft Report," U.S. Environmental Protection Agency, Air and
Radiation, EPA420-P-98-008, May 1998.
5. "Additional Study of Preconditioning Effects and Other IM240 Testing Issues,"
Prepared by Sierra Research for the U.S. Environmental Protection Agency,
Report No. SR98-02-01, February 2, 1998.
6. "Analysis of Alternate IM240 Cutpoints, Phase 2 Testing, and Exempting New
Vehicle Models on Test Duration and Projected I/M Benefits," Prepared by
Sierra Research for the Arizona Department of Environmental Quality, Report
No. SR98-05-01, May 12, 1998.
7. "Additional Study of Preconditioning Effects and Other IM240 Testing Issues,"
Prepared by Sierra Research for the U.S. Environmental Protection Agency,
Report No. SR98-02-01, February 2, 1998.
8. Ando, A., W. Harrington, and V. McConnell, "An Investigation of the IM240
Fast Pass - Fast Fail Algorithm," Resources for the Future, May 21, 1997.
9. Whitby, R., C. Shih, W. Webster, and R. Card, "Enhanced Inspection and
Maintenance Program: An Investigation of Real Time Data Utility and Dynamic
Quality Assurance Instrumentation - Task 1: A Predictive Modal Regression
Methodology for IM240 Fast-Pass/Fail Decisions," New York State Department
of Environmental Conservation, EPA Cooperative Agreement #X992008-01-0,
November 29, 1996.
10. "Additional Study of Preconditioning Effects and Other IM240 Testing Issues,"
Prepared by Sierra Research for the U.S. Environmental Protection Agency,
-67-
-------
Report No. SR98-02-01, February 2, 1998.
11. "Analysis of Alternate IM240 Cutpoints, Phase 2 Testing, and Exempting New
Vehicle Models on Test Duration and Projected I/M Benefits," Prepared by
Sierra Research for the Arizona Department of Environmental Quality, Report
No. SR98-05-01, May 12, 1998.
12. "Preconditioning Effects on I/M Test Results Using IM240 and ASM
Procedures," Prepared by Sierra Research for the American Automobile
Manufacturers Association, Report No. SR96-09-04, September 30, 1996.
13. "Additional Study of Preconditioning Effects and Other IM240 Testing Issues,"
Prepared by Sierra Research for the U.S. Environmental Protection Agency,
Report No. SR98-02-01, February 2, 1998.
14. Heirigs, P.L. and J. Gordon, "Preconditioning Effects on I/M Test Results Using
IM240 and ASM Procedures," SAE Paper No. 962091, 1996.
15. "Analysis of Alternate IM240 Cutpoints, Phase 2 Testing, and Exempting New
Vehicle Models on Test Duration and Projected I/M Benefits," Prepared by
Sierra Research for the Arizona Department of Environmental Quality, Report
No. SR98-05-01, May 12, 1998.
16. "Description and Documentation for Interim Vehicle Clean Screening Credit
Utility , Draft Report", U.S. Environmental Protection Agency, Air and
radiation, EPA420-P-98-008, May 1998.
17. "Analysis of Alternate IM240 Cutpoints, Phase 2 Testing, and Exempting New
Vehicle Models on Test Duration and Projected I/M Benefits," Prepared by
Sierra Research for the Arizona Department of Environmental Quality, Report
No. SR98-05-01, May 12, 1998.
###
-68-
-------
Appendix A
Regression Coefficients for Predicting Fast-Pass
HC, CO and NOx
1981-1985, 1986-1989, 1990 +
LDGV, LDGT1&2
IM147 Composite and IM147 Phase 2
-------
IM147 Composite Regression Coefficients, HC, 1981 to 1984 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.28302
0.49292
12.2873
2
0.22087
0.27692
3.2873
3.8277
3
0.20626
0.2453
2.3395
2.7004
3.2789
4
0.2025
0.23726
1.7016
2.6934
2.2595
1.8199
5
0.19674
0.23191
0.7267
2.1947
1.8508
0.9417
3.2711
6
0.19371
0.22842
-1.267
2.2279
1.7198
0.6659
2.2869
2.394
7
0.18738
0.21202
-0.6231
1.7266
1.5146
0.8648
0.7315
2.1029
3.5094
8
0.18284
0.20824
-0.687
1.7643
1.2784
0.7029
0.02
1.611
2.2764
2.4076
9
0.17769
0.19322
-1.3178
1.6712
1.053
1.0174
0.4016
0.8538
1.6285
2.0881
1.7321
10
0.13615
0.09383
0.4577
0.6228
1.0161
0.4833
1.0706
1.5597
0.8105
1.8333
1.0468
1.2416
11
0.09167
0.05089
1.0299
0.48
0.7316
0.8687
0.7352
1.3107
0.8971
1.067
0.9405
0.5907
1.4934
12
0.05768
0.02128
1.0603
0.5538
0.6507
0.8006
1.0811
0.9603
0.8179
0.8284
0.5147
0.7446
0.6189
1.5707
13
0.04274
0.01042
1.2346
0.5575
0.7831
0.7967
0.7346
1.1164
0.5348
0.8189
0.5691
0.7167
0.5419
0.9427
1.1583
IM147 Phase 2 Regression Coefficients, HC, 1981 to 1984 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.20703
0.20953
2.0143
.
11
0.13389
0.09921
.
,
.
0.8473
2.2185
12
0.08219
0.03676
1.0326
0.8734
2.2399
13
0.06134
0.01912
•
•
•
•
0.9799
0.7426
1.3775
1.5991
-------
IM147 Composite Regression Coefficients, HC, 1985 to 1989 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.2766
0.34656
13.6824
2
0.20812
0.16873
2.8674
4.3117
3
0.19563
0.14463
2.3164
3.1348
3.175
4
0.19112
0.1363
1.6523
3.1427
1.874
2.2098
5
0.18367
0.1339
1.0144
2.4615
1.5573
0.9572
3.7335
6
0.17993
0.13253
-0.7224
2.4585
1.4487
0.6521
2.5769
2.6014
7
0.17293
0.12064
-0.1377
2.0423
1.1284
0.6939
1.2277
2.0519
3.6615
8
0.16757
0.11845
-0.1218
2.0933
0.8965
0.3231
0.5267
1.6047
2.2931
2.5007
9
0.15987
0.10456
-0.8348
1.9171
0.7991
0.6696
0.7616
0.787
1.6851
1.8453
2.3036
10
0.11866
0.04344
0.6913
0.7331
0.8232
0.419
1.2369
1.4124
0.9527
1.7448
1.2797
1.2607
11
0.07898
0.02878
1.0914
0.5395
0.6123
0.8408
0.9244
1.1702
0.8734
0.9789
1.0522
0.5826
1.5578
12
0.05069
0.01474
1.0422
0.5726
0.666
0.783
1.1483
0.8874
0.779
0.8472
0.558
0.7298
0.6519
1.5595
13
0.03552
0.00709
1.2462
0.5894
0.7478
0.8463
0.7814
0.9755
0.537
0.8335
0.6423
0.7001
0.5807
0.8694
1.1763
IM147 Phase 2 Regression Coefficients, HC, 1985 to 1989 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.18785
0.12625
2.1945
11
0.11651
0.06121
0.8484
2.3783
12
0.07269
0.02771
,
.
1.0188
0.9382
2.2443
13
0.05129
0.01542
•
•
•
0.9743
0.8068
1.2835
1.6409
-------
M147 Composite Regression Coefficients, HC, 1990+ Model Year LDGVs
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C.12
C13
1
0.29616
0.11159
23.1504
,
,
,
,
2
0.20623
0.02228
5.2355
5.8375
3
0.18497
0.01433
3.0203
3.9074
4.8396
4
0.17964
0.01313
1.8314
3.998
2.6416
3.1051
5
0.16982
0.01882
1.4967
3.1511
1.9707
1.0296
4.4945
6
0.1626
0.02339
-0.6319
3.2159
1.5762
0.3816
2.8898
3.3072
7
0.15712
0.02176
-0.1996
2.6659
1.1993
0.6833
1.3686
2.6473
4.0134
8
0.15019
0.02365
-0.2736
2.8358
0.4878
0.4377
0.7262
1.815
2.3769
2.8785
9
0.14499
0.02087
-0.9649
2.6634
0.3657
0.612
0.8119
1.3348
1.9646
1.933
2.2046
10
0.11125
-0.00021
0.7575
0.9919
0.6501
0.5117
0.8424
1.9987
1.0142
1.6999
0.8989
1.6093
11
0.06826
0.008
1.2736
0.4331
0.8051
1.0733
0.6767
1.6782
0.6997
0.9987
0.5665
0.4893
1.8827
12
0.04475
0.0038
1.1284
0.4637
0.8565
0.8999
0.9022
1.2724
0.7224
0.801
0.5186
0.7095
0.67
1.6291
13
0.03163
0.00271
1.2701
0.5075
0.8315
0.9132
0.7195
1.199
0.6065
0.8076
0.5087
0.717
0.4785
0.9448
1.2534
IM147 Phase 2 Regression Coefficients, HC, 1990+ Model Year LDGVs
Mode
RMS
Reg
Regression Coel
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.18606
0.00868
,
.
,
3.3615
11
0.10709
0.01506
,
0.7724
2.9305
12
0.06909
0.00683
,
.
.
1.0539
0.9218
2.4854
.
13
0.04937
0.00481
•
•
•
•
•
•
1.0329
0.6164
1.4047
1.8847
-------
IM147 Composite Regression Coefficients, HC, 1981 to 1984 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.88112
1.20816
18.7688
2
0.55624
0.30403
2.7624
7.3858
3
0.53047
0.26019
2.1969
5.8824
3.2855
4
0.52156
0.25543
0.9977
5.8612
2.1358
1.8787
5
0.49019
0.25485
-0.044
4.8747
1.5253
0.0287
5.2326
6
0.48792
0.25374
-1.225
4.9215
1.5106
-0.3233
4.7417
1.1408
7
0.46241
0.21664
0.0211
3.9498
1.3363
0.3735
2.1518
0.4819
5.1699
8
0.44686
0.21504
-0.2613
3.9279
0.8207
0.4803
1.1191
0.1717
3.7016
2.687
9
0.42957
0.18296
-1.3774
3.5298
0.6848
1.0093
1.3393
-0.2624
3.2094
1.7057
2.5925
10
0.34155
-0.01373
0.8274
1.0373
1.0349
0.8487
1.8452
0.4581
1.4938
1.5469
1.9606
1.76
11
0.23673
0.02715
1.5626
0.4691
0.6981
1.2432
1.3615
0.6487
1.3509
0.9303
1.6045
0.4984
1.6815
12
0.14911
0.01514
1.4565
0.3769
0.7648
0.8643
1.4548
0.9108
0.8932
0.9314
0.5447
0.7877
0.503
1.7656
13
0.11909
0.02688
1.5445
0.4495
0.8326
0.9259
0.8438
1.1997
0.5083
0.8469
0.6296
0.6889
0.4349
0.9716
1.289
IM14
7 Phase 2 Regression Coefficients, HC, 1981 to 1984 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.58274
0.19292
3.7065
11
0.38837
0.1664
0.9113
2.7375
.
12
0.23224
0.06687
.
1.1491
0.6568
2.7135
13
0.18986
0.08178
•
•
•
•
0.9726
0.5315
1.5541
1.8213
-------
IM147 Composite Regression Coefficients, HC, 1985 to 1989 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.56663
0.7155
17.4831
2
0.41307
0.27941
4.1285
5.3274
3
0.38375
0.23937
3.4282
3.7248
3.8254
4
0.3776
0.23527
2.2862
3.6738
2.7612
1.9754
5
0.35141
0.23018
1.6416
2.7652
1.9068
0.203
5.4202
6
0.34487
0.2345
0.1878
2.7411
1.8426
-0.4005
4.4017
2.218
7
0.3303
0.20554
0.6469
2.1902
1.6331
0.1685
1.7815
1.6532
4.9504
8
0.32267
0.20591
0.7492
2.1797
1.2776
0.2305
0.9713
1.1658
3.8532
2.1259
9
0.31316
0.18966
-0.0959
2.0508
1.1976
0.5815
1.3116
0.6511
2.8444
1.3636
2.2753
10
0.2312
0.04245
1.3359
0.466
1.0324
0.3511
1.787
1.1352
1.8284
1.2853
1.5223
1.5724
11
0.15326
0.03569
1.5184
0.5314
0.3571
0.9645
1.1672
0.9774
1.4614
0.8891
1.4441
0.5349
1.5258
12
0.09268
0.01207
1.0718
0.5289
0.6221
0.8436
1.2171
0.8383
1.1437
0.7345
0.5506
0.7355
0.6079
1.6637
13
0.07251
0.00754
1.221
0.5562
0.7481
0.91
0.8623
0.9975
0.6153
0.7166
0.6573
0.6966
0.5766
0.9448
1.1525
IM14
17 Phase 2 Regression Coefficients, HC, 1985 to 1989 Model Year LDGT1&2S
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.38232
0.17582
,
2.907
11
0.2456
0.10791
,
.
0.9062
2.4096
12
0.13522
0.02745
»
,
.
.
1.0587
0.8344
2.4791
.
13
0.10561
0.01744
•
•
•
0.9906
0.7717
1.4444
1.6076
-------
IM147 Composite Regression Coefficients, HC, 1990+ Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.47358
0.21002
25.4779
4
,
.
2
0.30957
0.02101
4.8648
6.3816
3
0.27326
0.02516
3.888
3.781
5.3959
4
0.26274
0.02776
2.459
3.6512
3.7037
2.8503
5
0.24556
0.03553
1.5156
3.0487
2.2656
1.1333
5.1256
6
0.24105
0.04057
0.2926
3.0167
2.1731
0.3867
4.321
1.949
7
0.2245
0.03117
0.9559
2.2238
1.7687
0.8632
1.5799
1.1997
6.1164
8
0.21946
0.03551
0.8458
2.3821
1.239
1.018
0.8595
0.4381
5.2659
1.8798
9
0.20649
0.02736
-0.2205
2.1038
1.0238
1.4328
1.0755
0.1268
4.644
0.3345
3.178
10
0.14857
-0.0053
1.4529
0.2504
0.8221
0.9906
1.4828
0.8031
2.4979
0.7858
1.8619
1.8399
11
0.10014
0.00971
1.233
0.3838
0.75
1.249
0.8911
0.8007
1.2455
0.9788
1.3895
0.6068
1.5933
12
0.06561
0.00571
1.2678
0.3618
0.9336
1.0717
1.1601
0.6722
1.0448
0.7941
0.6573
0.7481
0.6237
1.5909
13
0.04847
0.00362
1.4218
0.4403
0.8949
1.0884
0.7455
0.8543
0.6521
0.8293
0.6849
0.7025
0.5718
0.8249
1.2326
1
M147 Phase 2 Regression Coefficients, HC, 1990+ Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.25439
-0.00921
3.7379
11
0.16351
0.01393
1.0922
2.61
12
0.10151
0.00637
,
1.1242
0.8847
2.4468
13
0.07579
0.00269
1.0318
0.7581
1.2923
1.813
-------
IM147 Composite Regression Coefficients, CO, 1981 to 1984 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
9.26688
12.0712
9.8906
2
7.04988
6.84257
4.2268
3.9173
3
6.25482
4.82892
1.9633
2.5947
5.6325
4
6.14698
4.60366
0.442
2.6858
4.0611
3.6215
5
6.03278
4.6086
0.2948
2.3842
3.4375
2.7805
2.445
6
5.9907
4.59648
-1.2868
2.4237
3.2458
2.053
2.0363
1.9446
7
5.88506
4.42293
-0.6192
1.9614
2.9124
2.029
1.1142
1.6931
2.3067
8
5.69421
4.31936
-0.5108
2.105
2.3926
1.0293
-0.2338
1.1296
1.7768
2.8351
9
5.48257
3.86232
-1.0847
2.0057
1.7045
1.8524
0.0911
0.1588
1.336
2.8233
1.2436
10
3.7416
1.41222
0.3751
1.1819
1.3859
0.4574
0.9912
1.3269
0.633
2.4857
0.6786
0.9644
11
1.73455
0.42403
0.5865
0.7047
0.7755
1.2595
0.9043
1.3168
0.5928
1.196
0.6803
0.6756
1.4246
12
1.02591
0.1691
0.6384
0.6819
0.7081
1.2496
0.8593
1.0366
0.6866
0.8982
0.631
0.7313
0.7978
1.1668
13
0.61707
0.07797
0.7481
0.6453
0.7525
1.1703
0.6936
1.1239
0.5095
0.9175
0.6368
0.7178
0.6804
0.7922
0.7957
IM
147 Phase 2 Regression Coefficients, CO, 1981 to 1984 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
6.2575
4.91001
1.5902
*
.
.
11
2.72138
1.37298
,
0.9487
2.1612
12
1.57636
0.56331
,
,
1.0279
1.1087
1.7699
13
1.04718
0.34246
•
•
•
•
•
•
0.9957
0.9216
1.2234
1.1113
-------
IM147 Composite Regression Coefficients, CO, 1985 to 1989 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
5.49052
7.60141
8.0871
2
4.401
4.59161
3.6606
3.4406
.
3
3.97643
3.43533
2.2184
2.3355
4.6236
4
3.89439
3.20531
1.0777
2.3748
3.2459
3.3247
5
3.82466
3.14867
1.1902
2.0936
2.7826
2 4739
2.174
6
3.78627
3.11501
-0.0293
2.1265
2.6293
1.7113
1.7618
1.9381
7
3.67681
2.8626
0.5686
1.7769
2.1844
1.6691
0.7635
1.5267
2.6631
8
3.60692
2.77387
0.6028
1.8114
1.9048
0.9633
0.1406
1.2651
2.1817
1.9243
9
3.39768
2.39443
0.1046
1.7076
1.3104
1.5167
0.5439
0.4973
1.7435
1.9067
1.2435
10
2.23224
0.99522
0.7545
1.0893
0.9857
0.8366
0.9455
1.4527
1.0902
1.9156
0.7336
0.8531
11
1.20706
0.33753
0.8722
0.7157
0.8315
1.123
0.8222
1.2181
0.7424
1.0531
0.6812
0.6913
1.3478
12
0.73525
0.12498
0.7748
0.6921
0.7973
1.0106
0.9017
0.8915
0.7162
0.8109
0.6412
0.7275
0.7963
1.206
13
0.41719
0.02516
0.8609
0.6739
0.7536
1.0237
0.7513
0.9565
0.5631
0.8254
0.6634
0.7132
0.6898
0.7947
0.8025
IM
147 Phase 2 Regression Coefficients, CO, 1985 to 1989 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Ticients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
3.68774
3.52859
,
1.3424
11
1.818
1.03471
.
,
0.9701
2.065
.
12
1.05353
0.3822
,
,
1.0173
1.1353
1.7852
.
13
0.60249
0.16835
•
•
•
•
•
•
•
0.9911
0.9591
1.1922
1.1246
-------
,
M147 Composite Regression Coefficients, CO, 1990+ Model Year LDGVs
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
5.62882
3.05114
14.1209
.
2
4.16263
1.40373
4.1999
5.0129
.
3
3.84928
0.90295
2.0243
3.5083
5.1234
4
3.79569
0.82093
0.6865
3.4697
3.4217
3.8707
5
3.74039
0.84809
0.9263
3.0501
3.0759
2.3548
2.4069
6
3.69164
0.84097
-0.418
2.9879
2.7724
1.4797
1.8325
2.8737
7
3.60802
0.72362
0.2888
2.5963
2.3389
1.3716
0.6475
2.02
3.1605
8
3.54512
0.73218
0.2011
2.6803
1.8672
0.4836
0.045
1.5147
2.5268
2.1618
9
3.3191
0.63591
-0.5877
2.5841
0.9946
0.913
0.3031
0.7605
2.0338
1.805
1.5015
10
2.33281
0.12368
0.8627
1.4067
0.7931
0.6929
0.5996
1.5372
1.2939
1.5496
0.7195
1.0174
11
1.14597
0.07195
0.8829
0.8553
0.8561
0.9971
0.6084
1.1998
0.5295
1.0312
0.6292
0.6836
1.427
12
0.64589
0.00678
0.7856
0.7476
0.8086
0.9368
0.7324
0.9804
0.6531
0.845
0.6498
0.7248
0.9011
1.0253
13
0.27203
-0.00513
0.7617
0.705
0.758
0.9566
0.6629
0.9463
0.5609
0.8841
0.6674
0.7136
0.7093
0.7561
0.774
IM147 Phase 2 Regression Coefficients, CO, 1990+ Model Year LDGVs
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
CI
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
3.72459
0.96776
1.7779
11
1.68958
0.30741
.
0.9902
2.121
12
0.94832
0.12176
1.03
1.2943
1.4878
13
0.43524
0.06182
•
•
•
•
•
•
1.0022
0.9992
1.0874
1.1022
-------
IM147 Composite Regression Coefficients, CO, 1981 to 1984 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
20.6965
24.9732
12.6811
.
2
13.3411
10.3032
4.0622
5.6571
3
12.3126
7.36435
1.776
4.3308
5.1971
4
12.1042
7.02622
-0.397
4.4381
3.5035
4.2351
5
12.0527
7.19205
-0.6418
4.2762
3.1383
3.695
1.3538
6
11.9685
7.20231
-2.5005
4.3477
2.8907
2.6546
0.8444
2.2266
7
11.807
7.08736
-1.7293
3.8076
2.5075
2.4323
-0.0037
2.0744
2.3528
8
11.6947
7.05512
-1.7922
3.9213
2.0893
1.8624
-1.1465
1.7128
2.0753
1.8569
9
11.3043
5.98779
-2.4399
3.5771
1.7377
3.1169
-0.5576
0.104
1.5691
1.7959
1.5672
10
8.49974
1.34387
-0.3377
1.8447
1.4618
1.2526
0.9093
1.3174
0.3783
1.7289
0.802
1.1466
11
3.6701
0.53011
0.3752
0.7417
0.797
1.456
0.6948
1.4668
0.637
1.0858
0.6348
0.6777
1.4646
12
1.98174
0.11753
0.7506
0.7285
0.5843
1.3695
0.8424
1.1166
0.6193
0.7984
0.5657
0.7829
0.8215
1.1544
13
1.09985
0.14803
0.816
0.6551
0.7798
1.2115
0.771
1.1356
0.4689
0.9171
0.5556
0.7201
0.6845
0.805
0.7714
IM147 Phase 2 Regression Coefficients, CO, 1981 to 1984 Model Year LDGT1&2S
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
13.6489
6.89656
2.104
11
5.65353
2.37259
,
.
0.9678
2.1519
12
3.07187
0.82239
1.0966
1.1263
1.7013
13
2.05416
0.91591
•
•
•
•
•
0.9926
0.9263
1.2294
1.0552
-------
IM147 Composite Regression Coefficients, CO, 1985 to 1989 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
ficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
13.8694
15.2236
13.6321
.
2
10.0033
6.93139
4.8288
5.0651
,
3
9.29649
5.09279
2.976
3.6934
4.9723
4
9.17411
4.84963
0.9452
3.7766
3.4685
3.917
5
9.01802
4.90791
0.7536
3.4092
2.7376
2.824
2.7994
6
8.92955
4.9686
-1.1289
3.4762
2.4874
1.4985
2.1637
2.6928
7
8.7691
4.69542
-0.4087
2.9985
1.953
1.7862
1.1736
2.1621
2.7068
8
8.59122
4.73532
-0.4517
3.0442
1.581
1.1186
-0.1135
1.556
1.9648
2.6236
.
9
8.26766
4.22168
-1.1718
2.6961
1.313
1.7854
0.4833
0.3603
1.3881
2.3245
1.6872
10
5.78009
1.23495
0.4919
1.2003
1.0982
0.5089
1.1372
1.5991
0.9224
2.0207
0.7215
1.1196
11
2.63622
0.43578
0.5368
0.7202
0.7365
1.367
0.8528
1.2188
0.6956
1.0567
0.6954
0.6738
1.4299
12
1.54921
0.09091
0.7468
0.6822
0.6828
1.1699
0.8768
0.9935
0.681
0.7599
0.6257
0.7538
0.8813
1.0928
13
0.75374
0.02099
0.8847
0.6423
0.7554
1.0555
0.785
1.0367
0.5445
0.8676
0.6146
0.7165
0.7117
0.7729
0.7783
IM147 Phase 2 Regression Coefficients, CO, 1985 to 1989 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
9.24585
4.98289
1.9147
.
.
11
3.97568
1.51254
,
.
0.971
2.1233
.
12
2.25662
0.44549
.
,
.
1.0591
1.2223
1.6066
13
1.25401
0.30824
•
•
•
•
•
0.9967
0.9833
1.1632
1.0721
-------
IM
147 Composite Regression Coefficients, CO, 1990+ Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
9.73928
4.40577
20.8542
.
2
7.10153
1.61757
4.8974
5.5895
3
6.61337
0.75824
3.3147
3.753
6.1468
.
4
6.57544
0.66445
2.1461
3.7345
4.9221
3.2434
5
6.35077
0.80324
2.4477
3.0432
3.6244
1.7761
4.4519
6
6.24116
0.84043
-0.0626
2.9872
3.2967
0.1221
3.6344
4.0729
7
6.18035
0.75951
0.6396
2.7006
2.7887
0.0988
2.4575
3.3995
2.5426
8
6.06991
0.85508
0.764
2.6575
2.3907
-0.3028
1.2626
2.5809
1.552
3.1236
9
5.74238
0.60774
-0.3973
2.3462
1.9317
0.2824
1.9415
0.7603
1
2.6996
2.1846
10
4.16315
-0.20486
1.0893
1.1803
0.9905
-0.2445
1.6449
1.896
0.6331
2.2387
1.0143
1.2424
11
1.8405
0.03592
0.7004
0.7222
1.0138
1.2204
0.7689
1.0977
0.5132
1.0686
0.7013
0.668
1.485
12
0.9886
0.02438
0.744
0.73
0.7776
1.1099
0.7872
0.5787
0.6457
0.7582
0.7113
0.7368
0.8642
1.1738
13
0.31426
-0.00518
0.7938
0.7228
0.7626
0.8887
0.727
0.7973
0.607
0.8159
0.6716
0.7166
0.7099
0.7665
0.7696
1
M147 Phase 2 Regression Coefficients, CO, 1990+ Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficaents
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
6.66055
0.6872
2.3246
.
11
2.67155
0.3375
,
.
0.9971
2.1714
.
12
1.38252
0.11966
.
1.0512
1.1996
1.6648
.
13
0.44345
0.05096
•
•
1.0135
0.9835
1.0996
1.0715
-------
IM147 Composite Regression Coefficients, NOx, 1981 to 1984 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Ticients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.88044
1.59831
13.9728
4
,
,
2
0.59034
0.7521
2.1278
4.9985
3
0.51405
0.55542
3.7485
3.858
6.0898
4
0.50608
0.53292
2.7857
3.7645
4.5611
4.8088
5
0.50038
0.52308
1.8882
3.3032
4.5487
3.9151
2.5841
6
0.49859
0.52116
-0.3119
3.3238
4.2688
3.1951
2.2305
2.3146
7
0.49285
0.51053
-0.1924
2.7423
4.4598
3.4302
0.6798
2.3994
2.7573
8
0.48391
0.49064
1.4802
2.7878
3.397
1.4246
-0.3484
1.2872
2.5791
2.9951
9
0.47485
0.4697
-1.3966
2.5877
3.4995
2.3294
-0.6719
-0.3573
2.0316
3.1392
1.9234
10
0.37891
0.23247
-0.0452
1.4342
2.2214
1.053
0.9414
1.3514
1.0033
2.6483
0.9386
1.2695
11
0.14698
0.05673
-0.0821
0.7178
0.8421
1.2211
1.0057
1.0475
0.6909
1.1895
0.5227
0.5818
1.6693
12
0.07162
0.02088
0.7266
0.73
0.768
0.8443
0.7672
0.6798
0.9669
0.7172
0.6657
0.7324
0.8746
1.0786
13
0.03296
0.00952
0.787
0.6373
0.9244
0.8663
. 0.7351
0.9267
0.6821
0.8366
0.5995
0.7084
0.7058
0.78
0.727
IM147 Phase 2 Regression Coefficients, NOx, 1981 to 1984 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Ticients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.62103
0.59564
.
2.552
11
0.21366
0.09595
.
.
0.8422
2.4327
12
0.09688
0.03109
.
.
.
1.0585
1.2349
1.5221
13
0.04004
0.01479
•
•
1.0012
0.9719
1.1811
0.9455
-------
IM14
17 Composite Regression Coefficients, NOx, 1985 to 1989 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.82399
1.36994
15.7886
.
2
0.54872
0.63552
3.0508
4.8202
3
0.48658
0.47811
3.9352
3.6357
5.939
4
0.47851
0.45047
3.6666
3.574
4.4109
4.4067
5
0.4678
0.4385
2.8415
2.9352
4.2742
3.2726
3.7315
6
0.46329
0.43505
-0.4665
2.9614
3.7155
2.6799
3.15
3.4084
7
0.45461
0.41619
-0.264
2.3408
3.9823
2.8116
1.3651
3.362
3.107
8
0.44117
0.3979
1.217
2.3973
2.7583
0.5364
0.1617
2.3482
2.8522
3.5198
9
0.43081
0.37005
-1.4203
2.2127
2.8121
1.3416
0.1059
0.4735
2.1588
3.7445
1.8009
10
0.34425
0.18854
0.3021
1.1817
1.6413
-0.2737
1.1837
2.0573
0.9764
3.1089
0.9139
1.2679
11
0.14249
0.05287
0.1216
0.6682
0.847
0.7423
1.0007
0.9561
0.7554
1.2453
0.5471
0.6313
1.6215
12
0.0723
0.01942
0.5926
0.7123
0.7055
0.7234
0.8115
0.6916
0.9342
0.7406
0.6586
0.7456
0.858
1.0946
13
0.02756
0.0071
0.618
0.6593
0.861
0.7847
0.7209
0.9137
0.6397
0.851
0.6173
0.7113
0.7061
0.7731
0.7423
IM"
47 Phase 2 Regression Coefficients, NOx, 1985 to 1989 Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.55293
0.44977
.
.
.
2.5168
.
11
0.20408
0.08145
0.9155
2.339
12
0.0984
0.02592
,
,
.
.
1.0681
1.2046
1.5383
13
0.03496
0.00939
•
•
•
•
*
•
0.9989
0.9746
1.1521
0.9785
-------
1
M147 Composite Regression Coefficients, NOx, 1990+ Model Year LDGVs
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
0.89663
0.69852
36.5998
2
0.51211
0.21316
5.1177
5.8472
3
0.44799
0.14368
6.3447
3.9439
7.1091
4
0.43718
0.11969
6.3263
3.8611
5.0794
5.5583
5
0.4116
0.11863
4.7571
2.675
4.7186
3.5952
6.8937
6
0.40287
0.11926
-0.6486
2.7059
3.7194
3.0461
5.7291
5.5197
7
0.39547
0.10825
-0.2279
2.0893
3.9596
3.32
3.74
5.0681
3.3616
8
0.37094
0.1121
1.8629
2.1003
2.1825
-0.0057
2.3861
3.5505
2.7425
5.2427
9
0.35919
0.09426
-1.8075
1.8994
2.0777
0.874
2.2073
1.6337
1.7328
5.4475
2.1483
10
0.27047
0.01275
1.1983
0.563
1.2164
-0.4082
2.2643
2.633
0.4833
3.3654
1.0148
1.6932
11
0.10736
0.01556
0.0524
0.6952
0.8359
0.782
0.8861
1.0436
0.6559
1.1881
0.5165
0.6625
1.6616
12
0.05405
0.00619
0.3724
0.7374
0.6968
0.7144
0.8282
0.6987
0.8659
0.7865
0.6759
0.7436
0.8588
1.0792
13
0.01741
0.0018
0.4251
0.6831
0.8395
0.7796
0.6819
0.8662
0.6071
0.851
0.6547
0.7112
0.7099
0.7711
0.7451
IM147 Phase 2 Regression Coefficients, NOx, 1990+ Model Year LDGVs
Mode
RMS
Reg
Regression Coe
ficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.42584
0.03497
,
.
3.2431
11
0.1545
0.02122
.
0.9719
2.3734
12
0.07551
0.00841
,
.
.
.
1.0818
1.2015
1.5197
13
0.02608
0.00273
•
•
•
•
•
¦
•
0.9985
0.9742
1.1431
0.9962
-------
IM147 Composite Regression Coefficients, NOx, 1981 to 19B4 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
Fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
1.77318
2.6756
22.2252
.
2
0.97763
0.81534
-1.0649
7.1138
3
0.8015
0.50589
2.4042
5.4388
7.4103
4
0.79236
0.48531
2.0196
5.3843
5.8747
4.4401
5
0.79068
0.48066
1.6893
5.1366
5.8394
4.1067
1.2009
6
0.78557
0.47575
-1.2448
5.1715
5.4814
2.6816
0.7411
3.3124
7
0.78143
0.476
-1.1956
4.6879
5.5481
3.0216
-0.6373
3.5383
2.1759
8
0.77522
0.46777
0.2093
4.7642
4.7945
1.4012
-1.6688
2.4918
2.0835
2.1312
9
0.75704
0.44065
-3.2536
4.3906
4.8993
2.5761
-2.1246
0.0658
1.5079
2.4101
2.4034
10
0.61652
0.14763
0.0597
2.7629
3.084
0.6688
-0.1034
2.211
0.8412
2.4747
0.7692
1.4734
11
0.23012
0.06297
-0.4836
0.9276
1.153
1.1834
0.6381
0.989
0.5432
1.561
0.4098
0.5401
1.7176
12
0.12238
0.03037
0.2166
0.7951
0.7934
0.968
0.7041
0.5695
0.9014
0.7627
0.6746
0.7513
0.9025
1.04
13
0.06163
0.01533
0.4239
0.6053
0.9275
1.0363
0.721
0.9598
0.6409
0.8842
0.6341
0.711
0.7041
0.7617
0.7441
IM14
7 Phase 2 Regression Coefficients, NOx, 1981 to 1984 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
1.14059
0.66921
,
,
,
3.6705
11
0.36161
0.1254
.
,
0.7893
2.5732
.
12
0.16442
0.03582
,
,
.
1.0922
1.2588
1.5043
.
13
0.0767
0.01903
•
•
•
•
•
1.0131
0.9198
1.2264
0.9462
-------
IM147 Composite Regression Coefficients, NOx, 1985 to 1989 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
1.53131
2.39066
19.1208
2
0.92778
0.80006
0.4734
6.3569
3
0.77287
0.47484
3.3615
4.7621
7.647
4
0.76394
0.4522
3.2324
4.7126
6.0512
4.47
5
0.75562
0.45153
2.6897
4.0983
5.8865
3.8567
2.922
6
0.74955
0.4519
-0.8381
4.1043
5.3452
2.7511
2.4874
3.6682
7
0.73604
0.43783
-0.7677
3.3026
5.526
3.2994
-0.0163
3.9319
3.8903
8
0.7273
0.42932
0.645
3.3545
4.6163
1.3597
-1.1299
2.8509
3.7622
2.6237
g
0.71142
0.3944
-2.8606
3.0614
4.7403
2.5337
-1.5906
0.297
3.1385
2.9429
2.1918
10
0.57504
0.1145
0.0543
1.5894
2.8092
0.5252
-0.1017
2.3307
1.9505
2.91
0.9769
1.4561
11
0.20555
0.04751
-0.4329
0.7123
1.0233
1.0206
0.7147
0.8005
0.794
1.4455
0.4808
0.581
1.7381
12
0.10175
0.02366
0.1643
0.7009
0.8486
0.8031
0.8477
0.5172
0.9143
0.6589
0.6773
0.7512
0.9067
1.0646
13
0.03924
0.00888
0.4472
0.6194
0.9382
0.8668
0.7417
0.8886
0.6209
0.8399
0.6328
0.7177
0.7075
0.7819
0.7267
IM147 Phase 2 Regression Coefficients, NOx, 1985 to 1989 Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.98114
0.5121
3.2652
11
0.30966
0.08548
,
0.8541
2.5226
.
12
0.14176
0.02736
,
,
.
1.0831
1.255
1.5035
.
13
0.05908
0.01277
•
•
•
•
•
1.0114
0.9489
1.2046
0.9338
-------
IM147 Composite Regression Coefficients, NOx, 1990+ Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
1
1.36435
1.26328
30.4323
2
0.68737
0.31269
2.0387
6.3658
3
0.58509
0.20124
3.9934
4.6556
6.8806
4
0.57465
0.17965
3.7451
4.6093
5.1071
4.9635
5
0.55221
0.18602
3.4381
3.6003
4.9732
3.6629
4.6002
6
0.54739
0.18882
0.7345
3.6265
4.3279
2.904
4.0176
3,3696
7
0.53147
0.17374
0.3162
2.8086
4.4764
3.2446
1.3899
3.6794
4.5038
8
0.51648
0.17351
2.2319
2.8117
3.2493
0.9649
0.0529
2.1267
4.2544
3.5644
9
0.50787
0.15312
-0.8203
2.6144
3.3792
1.6807
-0.2677
0.5626
3.5367
3.7081
1.6418
10
0.38986
0.0063
2.2347
0.9043
1.9881
0.5994
0.2319
1.5179
2.0767
2.4937
0.9883
1.6465
11
0.15532
0.01862
0.1437
0.6939
0.8694
1.0785
0.4364
0.5763
0.9011
1.13
0.5564
0.6361
1.7324
12
0.07617
0.00781
0.56
0.7
0.8067
0.8177
0.6048
0.4635
1.0357
0.6848
0.7114
0.7365
0.8807
1.0943
13
0.02687
0.00135
0.4313
0.6749
0.8709
0.7942
0.6777
0.8278
0.6267
0.8496
0.6436
0.7137
0.7073
0.7794
0.743
ir
k/1147 Phase 2 Regression Coefficients, NOx, 1990+ Model Year LDGT1&2s
Mode
RMS
Reg
Regression Coe
fficients
Number
Error
Constant
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12
C13
10
0.61055
0.04466
,
3.3215
.
11
0.22151
0.02607
0.9281
2.4391
12
0.10478
0.01138
1.053
1.2349
1.5211
.
13
0.03748
0.00291
•
•
•
•
•
1.001
0.9715
1.167
0.9652
------- |