-------
4. Data Description
From September 10, 1992 through March 19, 1993, EPA'3 contractor,
Automotive Testing Laboratories (ATL), conducted a vehicle testing program in
Mesa, Arizona, a suburb of Phoenix, on mostly 1983 and newer vehicles.
This program included several tasks designed to produce data for an
analysis comparing the ASM test to the ZM240 test as predictors of actual FTP
emissions. These tasks included the operation of an Arizona I/M inspection
lane. Vehicles at this lane received IM240s with a functional test of the
evaporative canister purge system (referred to as the purge test in the
remainder of this report), ASMa with the purge test, Arizona I/M tests, and
fuel system pressure tests under real-world I/M testing conditions. In
addition, vehicles were recruited from the I/M lane for additional tasks,
which included:
• FTP laboratory testing
• IM240 laboratory testing
• Contractor IM240-targeted repairs
• Commercial repairs obtained by vehicle owners to pass the official
Arizona I/M teat.
Choosing vehicles for laboratory testing was driven by the importance of
testing and assessing emissions from—and the impact of repair on—dirty in-
use vehicles. A random sample of vehicles visiting the I/M station would
result in the contractor recruiting mostly clean vehicles (see Section
5.2.5.8). But most excess emissions come from a relatively small percentage of
vehicles known as high to super emitters. To avoid the problem and cost of
evaluating a majority of vehicles that will ultimately be assessed as clean, a
stratified recruitment plan was employed to deliberately over-recruit dirty
cars, based on lane-IM240 results at the Mesa lane. A nominally 50/50 mix of
IM240-clean and DQ40-dirty vehicles were to be recruited for FTP exhaust
testing. In actual practice, nor* dirty cars than clean have been recruited
which is shown in Table 4.2.2.
Specifica_aonceraing the recruitment criteria and the teat procedures for
these task* are> discussed in Appendix A.
4.1 Data Listing*
Appendix B provide* a listing of the data used for the outpoint
effectiveness analysis* the contractor repair analysis, and the commercial
repair analysis, which are all discussed in Section 5.
13
-------
Data for the over 1400 vehicles that only received one set of lane tests
(no laboratory tests and no after-repair lane tests) are only available on
disk. These data include the purge analysis data, and the lane data used to
calculate ASM coefficients. The available disk(a) will include all IM240 and
ASM lane data including lane data for the laboratory tested vehicles. These
can be requested by contacting:
William M. Pidgeon
U.S. EPA
National Vehicle and Fuel Emissions Laboratory
2565 Plymouth Road
Ann Arbor, Michigan 48105-2425
Tel. No. 313-668-4416
Fax. No. 313-668-4497
Fax requests for data disks are preferred and a form is provided at the
end of Appendix B; questions can be addressed by phone.
4 .2 Database Statistics
The first official ASM/ZM240 test series was run on September 9, 1992.
Data collected up to March 17, 1993 were considered for these analyses.
During that period, 1574 vehicles received 1758 ASM/IM240 test series at the
Arizona I/M lane. Priority for testing was given to 1983 and newer model year
fuel injected vehicles. The following table illustrates the model year and
fuel metering distribution of the tested fleet:
14
-------
Table 4.2.1
Lana Data By Modal Year and Fual Mataring
MYR 1
81
82
83
84
85
86
87
88
89
90
91
92
To**1"
m 1
.
.
12
31
38
94
103
100
119
133
ISO
104
88S
TBI
.
2
6
30
42
S3
SO
61
77
48
35
11
415
ICARB
3
3
26
46
49
45
46
36
17
2
-
.
273
Totals
3
5
44
107
129
192
201
197
213
183
18S
US
1574
Of tha 1574 vahiclaa tastad 27 were recruitad for tha commercial rapair
program and 127 vahiclaa wara racruitad to tha laboratory for additional testa
and for contractor repairs whan tha rapair criteria tiara mat (Section S.6).
Tha following liat summarizes tha criteria used for recruiting laboratory
vehiclea and for data completeness:
• Tha IM240 and tha ASM wara designed to distinguish between
malfunctioning and properly functioning newer technology cars, so
only 1983 and nawar fuel-injected (no carbureted) cars wara used.
• One-half of the) laboratory vehicles were to exceed 0.80/15.0/2.0
(HC/CO/NOx) oa their laae-XM240.
• Ona-half of the) racruitad vehicles were to have tha lana-XM240
pexforaad price to tha ASM.
• Only vehiclea having aa as-received FTP, aa aa-received lane-IM240,
and aa as-recaived lana ASM teat wara used. Vahiclaa missing any one
of theaa three) taata wara not included in tha analysis.
Tha resulting database consisted of 106 fuel-injected. Table 4.2.2 lists
actual distribution statistics for these laboratory vahiclaa.
15
-------
Table 4.2.2
Distribution of Laboratory Recruited Vehicles
Lana-IM240
Passed: SO. 80 / 15.0 / 2.0
Failed: >0.80 / 15.0 / 2.0
Totals
Fuel
Metering
PFI
TBZ
PFI
TBZ
ASM Prior
to IM240
18
5
18
14
55
IM240 Prior
to ASM
14
3
23
11
51
Totals
32
8
41
25
106
Table 4.2.3 shows the model year and fuel metering distribution for the
106 laboratory recruited vehicles.
Table 4.2.3
Lab Data by Modal Tear aad Fuel Metering
MYR 1
83
84
85
86
87
88
89
90
91
92
Totals
PFI
6
12
7
13
9
5
7
6
7
1
73
1 TBI
2
8
7
6
1
4
2
2
1
.
33
Totals
8
20
14
19
10
9
9
8
8
1
106
Table 4.2.4 provide* FT* HC/CO emitter group statistics for these
recruited vehicles. FTP emitter groups are defined based on FTP emissions as
follows:
16
-------
I Normals! HC<0.82 and
CCX10.2
Highs :
0.82£HC<1.64
HC<1.64
and
or
and
CCX13 . 6
10 . 2SCO<13 . 6
Very Highs:
1.64SKX10.2
HCX10.2
and
or
and
C0<150
13.6SOX150
Supers :
HC210.2
or
CO2150
Table 4.2.4
Lab Vehicle FTP BC/CO Emitter Category Distribution
Ndramb
67
Highs
13
Very Hi«h
25
Supen
1
For more detailed information on the data used Cor these analyses refer to
Section 5. For details on the data excluded from these analyses refer to
Section 4.3 and Appendix C.
4.3 Quality Control (QC) Protocol
This Section provides a general description of the QC process. For more
detailed descriptions of the QC criteria and data excluded from these analyses
see Appendix C, which lists the QC criteria in detail and the vehicles removed
due to the QC protocol.
Data were received from ATT, in two forms. Calculated cycle-composite
values for all test* (lab and lane), except the ASM tests, and second-by-
second data for lane-IM240s and ASHs were provided. The calculated data and
the raw seeond-by-second data followed separate but similar QC processes. The
calculated data* were processed using a program which performed checks on FTP
data and 11424ft (lab and lane) data. These checks included bag result
comparisons/ fuel economy checks, test distance checks, dynamometer setting
checks, and test-to-test comparisons. For details on these checks see
Appendix C.
The second-by-second data were processed by a separate program which
performed similar checks for the raw data. The QC checks for the second-by-
second data included checks for acceptable speed, correct test/mode duration,
17
-------
sampling continuity, reasonable ambient background concentrations, acceptable
purge flow, and reasonable fuel economy. Again details concerning these QC
criteria are included in Appendix C. The second-by-second QC program also
calculated composite values for the IM240 and ASM tests.
In addition to the QC program comparisons, the calculated results reported
by ATL were compared to those results calculated from the second-by-second
data. Significant differences were investigated. All lab vehicles violating
the QC criteria were hand checked by EPA staff and the data were corrected or
removed, as warranted. Due to the volume of lane data, lane vehicles that
violated QC tolerances were removed from all pertinent analyses, without
further attempts to "save" the data unless solutions were obvious. These
unutilized data will be checked, as time permits, for future use. In
contrast, because the vehicles that received FTPS were relatively precious,
significant effort was expended to correct data that were identified by the QC
process.
Vehicles removed from the sample are discussed in Appendix C on page C-4
through C-8.
18
-------
5. Analyses/Discussion
5.1 Introduction
The purpose of Section 5 is to present the analysis EPA used to assess
whether the ASM test ia sufficiently effective in identifying high emitting
cars needing repair when compared to the ZM240 test, and the findings of that
analysis. Additionally, it provides a comparison of the repair issues for
those vehicles that were identified aa needing repairs.
Section 5.3 compares the ability of the ASM and the IM240 to identify
vehicles needing repair, and presents EPA'a major findings regarding the
effectiveness of the ASM aa an alternative to the IM240 for enhanced I/M
programs. It discusses comparisons of IM240 versus ASM using information from
cutpoint tables. Section 5.2 provides information needed to understand how
the cutpoint tables were derived.
Section 5.4 compares the correlation of the IM240 and ASM with the FTP
using traditional regression analysis. Section 5.5 discusses the somewhat
specialized issue of how four ASM scores are combined in one score and the
uncertainties and sensitivities in this process.
Section 5.6 discusses the repairs performed by the contractor and repairs
performed by commercial repair shops.
Section 5.7 discusses canister purge system test results and Section 5.8
discusses methods that will be explored to improve the power of the IM240.
5.2 Analyse* Techniques
This section discusses the methodology and criteria EPA used to compare
the ability of the ASM and the IM240 to identify vehicles needing repair.
This section explains why the criteria are important, how the criteria are
derived, and indicates the tradeoffs associated with these interrelated
criteria. Than* Section 5.3 contrasts the ASM and the IM240 using the
criteria explained ia Section 5.2.
5.2.1 Reducing Four Steady-State Modes to e Single
Score per Pollutant roc Comparison, to One
Cutpoiat per Pollutant
This section explains how the final ASM score is computed. Two questions
will be answered ia this section:
19
-------
1. Should the four mod* scores for each pollutant be combined to
calculate a single result or score for each pollutant, or should a
separate score be reported for each of the four modes, and apply
those separate scores to separate outpoints for each of the four
modes?
2. How is the score computed for each ASM mode?
5.2.1.1 Reporting Overall ASM Results Versus Reporting
Individual Mode Results
There are three alternatives for reporting overall ASM test results. The
first alternative does not combine the scores from the separate modes, so this
alternative is analogous to the way 2500/idle test results are reported. The
HC and CO scores are reported for the 2500 mode and separate HC and CO scores
are reported for the idle mode for a total of four scores and up to four
outpoints. For the four mode ASM test, this is too complicated. With NOx
added, three outpoints are needed for the 3 dynamometer modes and two
outpoints (HC and CO only) for the idle mode, necessitating 11 separate
outpoints. (Because NOx emissions are insignificant during an idle test, NOx
is only considered for the 3 dynamometer modes.) This first alternative is
too unwieldy for a four mode test.
The second and third alternatives are two different ways to report a
single score for each pollutant by combining one pollutant's scores from all
the modes.
For the second alternative, the single score would be the sum of the
scores from each mode, using a weighting of 25% for each mode. For example,
to calculate the single ASM score for HC, the equation would be as follows:
ASM HC »(0.25 * 3015 HQ + (0.23 * 2525 HQ + (0.25 » 50RL HQ + (0.23 * idle HC)
In the third alternative, which EPA used, a single score is determined
from the sum of the individual mode scores, but the weighting or coefficient
for each was determined by regression techniques. A multiple regression was
performed wherein all four of the mode scores are independent variables that
were regressed against FT* scores. The regression produced coefficients for
each mode, plus a constant. These coefficients weight each mode more
appropriately than the second alternative's method of just assigning each mode
a weighting of 25%. BAR/Sierra used this regression method, and likewise
EPA's analyses for this report also used this regression method, with one
difference that is discussed in Section 5.5. This yields an equation to
20
-------
calculate a single ASM score for each pollutant. For example, the equation
for calculating a single ASM HC score is as follows:
ASM HC - (x * 5015 HQ + (y * 2525 HC) + (z * 50RL HQ + (t * idle HQ + Constant
the x, y, z, t, and constant terms are listed in Table 5.2.2.
While ASM advocates have used this concept, none have proposed specific
coefficients for EPA to evaluate. Thus, EPA had to develop coefficients.
The remaining question is: "How were each of the individual mode scores
determined?"
5.2.1.2 Determination of Individual Mode Scores
Emission concentrations were measured on each of the four ASM modes (see
Section 5.2.3 for more details). These concentration measurements were then
converted to simulated grams/mile emissions, because concentration
measurements do not provide a reliable indication of the magnitude of
pollutants emitted per mile traveled. At the same exhaust concentration
level, a heavy vehicle will emit more per mile than a light vehicle.
To calculate simulated g/mi results, EPA followed BAR/Sierra's method,
which was also followed by ARCO, wherein the measured concentration values are
multiplied by the Inertia Weight (engine displacement for the idle mode) of
the vehicle. ttie Idle Mode was not considered tor NO* since it is a no load
test. EPA also divided these simulated g/mi results by the scaling factors
listed in Table 5.2.1. Using these factors yield overall ASM scores that have
magnitudes similar to FTP and IM240 magnitudes.
Table) 3.2.1: feeling r act ore Used to Keep
Regression Coefficients of Bqual Magnitude)
Pollutant
HC
CO
NOx
Modes 1-3
rcoNci * » / x
10*
10'
10<
Mode 4
[COMC1 * DiSP(L) / X
103
10°
NA
5.2.2 Multiple Linear Regressions to rind
Coeffioienta
As previously discussed* multiple linear regressions were performed using
the four modes (three for NOx) of the ASM test as the independent variables
21
-------
(XI,...,X4) The one difference mentioned in Section 5.2.1.1 above is that the
IM240 (rather than the FTP) was used as the dependent (Y) variable*. This was
done for testa on which the ASM was run first only, because the corresponding
IM240s are pre-conditioned, and thus more closely resemble an FTP.
Vehicles that were recruited to the lab or received commercial repairs
were not included in the database used to develop coefficients, because these
are the cars to which the coefficients were applied. EPA determined that
including these would cause the developed coefficients to mask the test
variability of the ASM. (This is also discussed in Section 5.5.)
The multiple linear regressions were run on a database of 608 lane ASM
tests versus pre-conditioned lane-IM240s, giving the following coefficients
for each mode.
Coefficient*
Table 5.2.2
Developed from Multiple Regression ASM
(see Table 5.2.1 for scaling factors)
Versus IM240
Mode
Constant
5015
2525
50MPH
Idle
Adjusted Rz
HC
0.083
0.025
O.OS9
0.136
0.124
29.0%
CO
2.936
0.040
0.043
0.356
1.350
50.1%
NOx
0.258
0.061
0.219
0.352
NA
59.1%
5.2.3 Applying ASM Coefficients
The coefficients were then used to calculate composite ASM scores for all
lab vehicles and coonercially repaired vehicles. These are the ASM scores
that arei reported in the. ensuing outpoint tables, scatterplots, and
regression*.
* Why the XM240 waa used aa the dependent variable, rather than the FTP, is
explained in Section 5.5. Thia ia not discussed here because the purpose of
this section ia to explain how, rather than why. Also, this issue requires a
lengthy discussion and reliea on information presented in Section 5.5, so
repetition ia also avoided.
22
-------
5.2.4 ASM Concentration Measurements
The ASM concentrations were measured over a 40 second period. Because the
exhaust sample delay to the most downstream analyzer cell is almost 10
seconds, the first 10 seconds of data were ignored. The concentrations that
are used in the composite ASM score calculations are actually reported
averages over the last 30 second period. For various reasons, the time
allotted for measured concentrations was occasionally less than 30 seconds.
In these few cases, EPA calculated the average concentrations over this
shortened period and reported these values. No ASM tests were accepted with
concentrations averaged over a period of less than 20 seconds.
5.2.5 Explanation of the Criteria Used To Compare I/M
Tests
In assessing the overall effectiveness of the ASM relative to the ZM240,
it is important to determine their effectiveness in measuring and determining
a variety of factors, including the excess emissions identified, the failure
rate, the error-of-commission rate, the two-ways-to-pass criteria, the
discrepant failures, and the, unproductive failure rate. Each of these is
discussed below. These criteria are used in Section 5.3 to compare the
effectiveness of the two procedures.
5.2.5.1 Excess Emission Identification Rate (ID*)
EPA commonly uses the rate of excess emissions identified during an I/M
test to objectively and quantitatively compare I/M test procedures. Excess
emissions are those FTP-measured emissions that exceed the certification
emission standards for the vehicle under consideration. For example, a
vehicle certified to the 0.41 g/mi HC standard whose FTP result was 2.00 g/mi,
would have excess emissions equalling 1.59 g/mi HC (i.e., 2.00 - 0.41 - 1.59).
The excess emissions identification rate (IDR) equals the sum of the
excess emissions for thsj vehicles failing the I/M test divided by the total
excess emissions) (because* of imperfect correlation between I/M tests and the
FTP, some I/Mfpsssing vehicles also have excess emissions which are used for
calculating thei total excess emissions). Thus, assuming an I/M area that
tests 1000 veaieles, 100 of which arc emitting 1.59 g/mi excess emissions
each, while the) I/M test, fails (identifies) 80 of the excess emitting
vehicles, the excess emission identification rate can be calculated as
follows:
8Q failing vehicles * 1.59 a/m\ excess per vehicle A an.
100 vehicles * 1.59 g/mi excess per vehicle
23
-------
EPA uses IDR instead of merely comparing the number of vehicles that
correctly fail and correctly pass. The IDR better contrasts the relative
merits of competing I/M test procedures because failing vehicles with high
emissions is more important than failing those that are only slightly above
their certification standards. For example, take two I/M procedures that
correctly failed 100 of the 500 vehicles that had FTP emissions greater than
their certification standards, but only SO cars failed both tests. If the
fifty cars that failed Test A were high FTP emitters, and the other 50 cars
that failed Test B had FTP emissions only slightly above their standards,
obviously Test A would be preferred, and its IDR would reflect its better
performance. Test A's better performance is not evident in comparing the
number of vehicles that correctly fail.
5.2.5.2 Failure Rate)
As the IDR increases with different test procedures or different
outpoints, the opportunity to identify vehicles for emission repairs also
increases. However, this measure is not sufficient for determining which is
the more efficient and cost-effective I/M test. Other criteria must also be
addressed before such an assessment can be made. One such criterion is the
failure rate, which ia calculated by dividing the number of failing vehicles
by the number of vehicles tested. For example:
50 vehicles failed I/M ., ,A_ «•.,/«* ji
1000 vehicle. tested * l°° " 5% 1/M f*ilur* rat-
The ideal I/M teat ia on* that fails all off the dirtiest vehicles
while passing those below the FTP standard or close: to it but still above it.
The potential emission reduction benefit decreases as emission levels from a
vehicle approach the standard, because the prospect for effective repair
diminishes. Thus, achieving a high IDR in conjunction with a low failure rate
(as a result of identifying fewer vehicles passing or close to the standard)
efficiently utilise* resources.
5.2.9.3 Brxos-o£-Commission (Bo) Rat*
Properly functioning vehicles which pass FTP standard* sometimes fail the
I/M test; theae are referred to aa false failurea or errors-of-commission
(Bcs). when error-of-commission vehiclea are sent to repair shops, no
emission control system malfunctions exist. Often, the repair shop finds that
the vehicle now paasea the teat without any change*. Theae false failures
wast* resources, annoy vehicl* owner*, and may lead to emission* increases as
a result of unnecessary and possibly detrimental "repairs." Automobile
manufacturers see this aa a significant problem, sine* it can contribute to
24
-------
customer dissatisfaction and increased warranty coats. An I/M program seeking
larger emission reductions through more stringent emission test standards may
actually increase the number of false failures. The error-of-commission rate
is, therefore, an important measure for evaluating the accuracy of I/M tests.
To see how an error-of-eommission rate is calculated, assume an I/M area
which tests 1000 vehicles, of which 100 fail the I/M test, although only SO of
those 100 failing vehicles also exceed the FTP standards. The error-of-
commission rate equals the number of vehicles that fail the I/M test while
passing the FTP divided by the total number of vehicles which were I/M tested:
50 vehicles failed I/M but passed FTP „ ,AA c. _
100Q v.hicl.3 * 100 - 5% EC rat.
As the error-of-commission rate decreases, vehicle owner satisfaction and
acceptance of the I/M program increases. Thus, while it is relatively easy to
improve the IDR by making the I/M test standards more stringent, this
*improvement* comes at the cost of potential increases in the error-of-
commission rate.
5.2.5.4 Two-Ways-To-Pass Criteria
The theory behind the two-ways-to-pass criteria for the IM240 is as
follows. Assuming that the IM240 test was correctly performed in the first
place, the most likely reason that a properly functioning vehicle would fail
an IM240 is that the evaporative canister was highly loaded with HC molecules
and that they were being purged into the engine during the test. This has
been a significant cause of false failures in existing I/M programs and it has
been shown that highly loaded canisters can cause both high HC and CO
emissions, even though the feedback fuel metering system is functioning
properly.
Since the canister is being purged during the IM240, the fuel vapor
concentration fro* the canister continually decreases during IM240 operation.
The decreasing; fuel vapor concentration results in decreasing HC and CO
emissions. So£ emissions during Mode-2 (the last 136 seconds of the 239
second cyclep should be) lower than the composite results. On the other hand,
if the vehicle is actually malfunctioning, Mode-2 emissions should remain
high.
Catalyst temperature can also affect test outcome. Emissions are
generally highest after a cold start, before the catalyst has had a chance to
warm up. If a vehicle is standing in line for a prolonged period of time, or
was not sufficiently warmed up before arriving at the test lane, this can
25
-------
cause the vehicle to fail, whan, in fact, it should be passed. Under the two-
ways-to-pass criteria, Mode-1 acts as a preconditioning mode, thus providing
insurance against this particular variety of false failure.
NOx outpoint criteria are not included in EPA'3 two-ways-to-pass
algorithm. So a vehicle which meets the ZM240 NOx outpoint (i.e., composite
NOx £ 2.0) only fails if both its composite emissions exceed the HC or CO
composite outpoints, and its Mode-2 emissions exceed the HC or CO Mode-2
outpoints. In other words, a vehicle can pass by having low HC/CO emissions
in Mode-2 even if its Mode-1 HC/CO emissions were high. EPA is mandating this
approach to IM240 outpoints.
The ZM240 outpoint tables, in Appendix E and Table 5.3.1 in the next
section, were calculated using the two-ways-to-pass-criteria.
The two-ways-to-pass criteria were optimized only at the outpoints EPA
recommends for biennial enhanced Z/M programs, which are referred to as
"standard* or "recommended* ZM240 outpoints. For composite emissions, the
standard outpoints are 0.80 g/mi HC, 15.0 g/mi CO and 2.0 g/mi NOx. The Mode-
2 criteria for the standard outpoints are 0.50 g/mi HC and 12.0 g/mi CO. The
Mode-2 outpoints were carefully selected from EPA'a ZM240 data collected in
Indiana, to pass properly functioning vehicles while continuing to fail
malfunctioning vehicles. (The Mode-2 criteria were not redetermined for this
new Arizona sample.) The) Mode-2 criteria, listed in the outpoint tables in
Appendix r and Table 5.3.1, simply increase proportionally with increasing
composite outpoints (i.e., become less stringent) and decrease proportionally
with decreasing composite outpoints (i.e., become more stringent). The point
is that these Mode-2 criteria have not been optimised at every stringency
level to provide the) beat tradeoff among IDR, failure rate, and Ecs, so it is
probable that the effectiveness of the ZM240 Mode-2 outpoints can be improved.
5.2.S.S Discrepant Failures (DFs)
Discrepant failure* are vehicles that fail an Z/M teat for HC and/or CO
and pass the> ffs? foe HC and CO, but fail the FTP for NOx, or vice versa. The
table below illustrates one) possible discrepant failure scenario:
Teat
Short Teat
FTP
HC or CO
Pass
Fail
NOX
Fail
Pass
Zn thia example, a false failure for NOx happens to occur on a vehicle
which is a false pass for HC/CO.
26
-------
Repair diagnostic routines are frequently selected on the basis of which
pollutant caused the I/M test failure. Given that HC/CO and NOx move in
opposite directions with changes to the A/F ratio, there is not much reason to
expect that fixing a NOx problem will reduce HC/CO emissions. Therefore,
these scenarios represent an error of sorts for the short test. If a vehicle
was to fail the short test for NOx, whereas the only high FTP pollutant was
CO, chances are the mechanic will be looking for a problem that causes high
NOx. In this case, the problem that is causing high CO emissions is not
likely to be found.
5.2.5.6 Unproductive Failure (UP) Rate
The unproductive failure rate represents the percentage of vehicles that
will be identified as needing repair, but either repair is not needed (Ecs),
or it is not likely the reason for repair will be found (DFs). The
unproductive failure rate is calculated by adding errors-of-conmission to
discrepant failures, and dividing the quantity by the total number of vehicles
which were I/M tested. Keeping with the same example as above, take an I/M
area which tests 1000 vehicles. 100 fail the I/M test, 50 of those 100
failing vehicles are Errors-of•Commission, and S are Discrepant Failures:
50 Bea + 5 DFs » ,«« . ... .,_.*
1000 vehicle, tested * 10° ' 5 5% OT rat«
Unproductive Failure
5.2.9.7 Vehicles with Malfunctions that Were Mot
Counted a* EC* and DFs
Errors-of-commission in I/M programs have been most often caused by test-
to-test variability or incompatibility between the I/M test procedure and
vehicle emission control system* (e.g., air pump switching), so attempting to
repair Ee vehicles) were) fruitless. With the IM240, however, EPA has found
that some vehicle* that had failed the ZM240 and passed the FTP actually did
have malfunction*, so they were correctly identified and air quality would
suffer by ignoring the*). By the strict definition of Ecs, the IM240 is
penalized despite it* successfully identifying malfunctioning vehicles.
A likely reason for vehicle* passing the FTP despite a malfunction is
that malfunction* are sometime* intermittent, vehicle 3172 provides a good
example. This vehicle had a number of IM240s performed, some with high NOx
and others with low NOx. The mechanic indicated that the vehicle had a sticky
EGR valve. The mechanic's diagnosis was not influenced by the FTP result
because the contractor had been instructed to report only IM240 scores to the
27
-------
mechanics, not the FTP score. This has become standard practice to allow the
contractor-repairs to simulate commercial repairs/ where mechanics will not
have access to FTP results.
For this analysis, EPA did not count vehicles as Ecs when they had a
malfunction that would logically explain the IM240 test failure. To
facilitate a fair comparison between the ASM and the IM240, the ASM failing
vehicles that passed the FTP, but had malfunctions, also were not counted as
Ecs when their malfunction would logically explain the ASM failure.
EPA was very conservative in that a vehicle was counted as an EC unless
the malfunction clearly explained the ASM or IM240 test failure. For example,
vehicle 3239 failed the IM240 with 2.4 g/mi NOx yet passed the FTP. The
vehicle was diagnosed as having a slow responding 02 sensor and it was
replaced. Because (1) a report of a slow-responding 02 sensor does not
indicate that objective criteria were used, (2) NOx failures are not strongly
associated with defective 02 sensors, and (3) all of its other IM240 tests
had passing NOx, the car xaa. counted as an EC despite the mechanic's judgement
the the O2 sensor should be replaced, which he did.
Using the similar logic, some vehicles with discrepant failures were also
not counted aa DFs when their malfunctions could logically explain the I/M
test failure and a proper repair could be expected to reduce FTP emissions of
the affected pollutant even though FTP emissions of that pollutant were
initially below FTP standards. For example, the vacuum leaks on vehicle 3154
could cause a lean air/fuel ratio which can lower the catalyst's NOx
conversion efficiency and cause higher combustion temperatures, both of which
can causa high NOx on the ZM240 and ASM. FTP NOx emissions should also be
affected but perhaps not enough to cause an FTP failure. Because it is
logical for a mechanic to check for vacuum leaks on a car that fails NOx, and
this vehicle did have vacuum leaks, the I/M tests shouldn't be penalized for
correctly identifying the malfunction, on the other hand, if this vehicle had
failed CO on aa I/M test and NOx on the FTP, the mechanic would look for
problems causing a rich air/fuel ratio, which would probably preclude looking
for vacuum leak*.
Table 5.2 list* the five vehicles the met the strict definitions for Ecs
or DFs, but were not counted for the reasons discussed. Note chat while these
vehicles were not counted a* Scs or DFs in the outpoint tables, they still do
count toward the failure Rate.
28
-------
Table 5.2.3.1: Cars not Counted •• «cs or OWm
Vehicle
Original Status
Malfunctions Explaining
I/M Test Failure
3154
3172
3200
3216
3244
Discrepant failure; failed
IM240. and ASM NOx, but
failed FTP CO only.
Error-of-Coaudaslon; IM240
NOx.
Discrepant Failure; failed
IM240 and ASM NOx, but
failed FTP HC only.
Injector seals leak at
intake manifold;
distributor advance vacuum
hose broken.
EGR valve sticks, EGR valve
vacuum line plugged.
EGR position sensor out of
range.
Brror-of-Commission/ ASM HC ECM malfunction
£rror-o/-Commission/ XM240
and ASM NOx.
Injector malfunctions
intermittently.
5.2.5.• Weighting Factor* to Correot Biased Recruiting
The criteria used to recruit vehicles for laboratory testing heavily
biased this laboratory sample toward ZM240 failing vehicles. Sixty-two
percent of the 106 laboratory vehicles had failed the lane-!M240 criteria
(>0.80/15.0/2.0), whereas only 19% of 2,070 cars tested at the Ian* failed the
IM240. This resulted in a laboratory sample that was highly biased toward
failing vehicle*. (Two-ways-to-pass criteria was not considered for
laboratory recruiting.)
Using this biased database results in unrealiatically high excess emission
identification rates, and unrealiatically low error-of-commission rates. So
the laboratory database- most be corrected to represent the pass/fail vehicle
ratio in the*ia-ose fleet to correctly determine iDRs, failure rates, and Ecs.
The database- was) corrected using the weighting factors presented in Table
5.2.5.2.
Weighting factors are used as follows: If the 66 failing vehicles that
received FTP tests had excess HC emissions which totaled 100 g/mi, the
database would be corrected in this case by multiplying 100 by the 5.97
weighting factor, resulting in corrected total excess emissions of 597 g/mi
for the dirty vehicles. In comparison, the total excess emissions of the
29
-------
IM240-clean vehicles have to be multiplied by 41.9 to make their excess
emissions representative. The total simulated excess emissions are the sum of
the simulated excess emissions from the clean and dirty vehicles in the I/M
lane sample. The number of vehicles tested was similarly adjusted with the
factors for the purpose of calculating failure rates. The sample of 40 clean
vehicles provides confidence in conclusions about a teat's relative tendency
to avoid failing clean cars.
Table 5.2.5.2
Weighting raetors Used To Adjust the Laboratory Database
IM240 at Lane
Paaa:
rail:
SO. 80/15. 0/2.0
>0. 80/15. 0/2.0
t at Lane
1676
394
f at Lab
40
66
Weighting Factor
41.90
5.97
The resulting weighted database was used to produce the realistic
estimates of IDRa, failure rates, and Ecs that are listed as outpoint tables
in Appendices D 4 E. These outpoint tables are sorted by failure rates. For
the outpoints that produce the same failure rate, the results are sorted first
by HC IDRs (in descending order) and then by NOx IDRs.
30
-------
5.3 Comparison of XM240 Versus ASM Using Cutpoint Tables
In assessing the overall effectiveness of I/M test procedures, as
discussed in Section 5.2, it is important to determine the test's
effectiveness in terns of IDR, the failure rate, discrepant failures and
unproductive failure rate.
Appendices D and E list the same criteria for many different outpoints.
Table 5.3.1 provides a summary of these criteria to compare the ASM with the
IM240 for the following three important scenarios:
- ASM outpoints selected to achieve the same 18% failure rate (using
the outpoint tables that are reweighted to correct the lab sample
bias) that result from EPA's recommended IM240 two-ways-to-pass
outpoints of .80/15.0/2.0 + 0.50/12.0. Among the ASM outpoint
combinations with this failure rate (see Appendix B), a combination
was selected that produced the maximum IDRs for all the pollutants
simultaneously, so there was no need to set priorities among
pollutants.
- ASM outpoints selected to achieve IDRs similar to the IDRs that
result from EPA's recommended IM240 two-ways-to-pass outpoints of
.80 / 15.0 / 2.0 + 0.50 / 12.0. Because ASM CO and NOx IDRs could
more favorably be presented by excluding HC, two ASM outpoint sets
are presented* on* to provide matching ASM and IM240 HC IDRs
(resulting in better IDRs for CO and NOx), and the second to provide
matching ASM and IM240 CO « NOx IDRs.
- ASM and IM24O outpoints selected to achieve th* highest IDRs possible
whil* keeping th* unproductive failure rat* below 5%. This case was
addressed oa th* possibility that an aggressive I/M program might be
willing to op*rat* with such a high Be rat*.
31
-------
Table 5.3.1
Comparison of th* Ability of IN240 and ASM to Identify Vehicles
Whoa* Emission* Exceed Certification standards Baaad on 106 Lab
Vehicle* Weighted to Represent 1676 Lan* Vahiclaa
Eicin Emissions
Failure Identified
Test Scenario Rate HC CO NOz
% % % %
IM240 Staadnd IS 92.2 67.5 83.4
Ctpcs.
ASM SaaMFafl IS 74.7 61.1 68.0
Rat*
ASM SimilatHC 42 92.4 78.1 95.0
IDR
ASM Similar CO* 24 80.4 66.2 89.4
NOx IDR.
ASM BeetlDRa 2S 82.5 67.0 80.1
w/UF9<3%
IM240 BestlDb 33 95.9 78.2 97.1
w/UFe<5%
Weighted f
ofVehielaf 1676
Unproductive,
Discrepant Failure
EC* Failures Rate** Cutpotnts
# # %
0 12 0.6 .80/15.0/2.0
+
0.50 / 12.0
42 6 2.3 1.00/8.0/2.0
174 180 17.1 1.00/8.0/1.0
84 48 6.4 1.00/11.0/1.4
48 48 4.6 .40/8.0/1.5
60 12 3.5 JO / 9.0/1.7 +
.19 / 7.0
* Exclude* Bo vahlclaa that had malfunctions that cauaad an I/M tast failure,
but bacauao thoy wr« intaraittant malfunctions, did not fail th« FTP. FTPs
war* always performed on a diffarant day. Sine* thay w*r* corractly
idantifiad by th* I/M t**tf they ar* not vahiclaa that will "ping-pong".
** The Unproductive Failure Rate includes th* traditional EC vehicles and the
discrepant failures, without including th* traditional Be vehicles that were
found to have intermittent malfunctions that were not identified by th* FTP
test.
32
-------
For the first scenario with an 18% failure rate for both tests, the ASM
statistics in Table 5.3.1 show that the IM240 identifies 18 percentage points
more of the excess HC emissions and 15 percentage points more of the excess
NOx •missions than the ASM identifies, with a significantly lower unproductive
failure rate. Expressed differently, an lM240-based program would relinquish
about 19% of its HC effectiveness and 18.5% of its NOx effectiveness by
substituting the ASM test at the same failure rate. Some relatively dirty
vehicles are missed by the ASM and replaced by relatively clean vehicles.
This scenario is illustrated in Figure 5.3.1.
The second scenario shows that in order to achieve HC IDRs similar to the
IM240'3 at an 18% failure rate, the ASM'a failure rate must be increased to
42%, resulting in an unacceptable EC rate of 17%. To achieve similar CO and
NOx IDRs with the ASM, an ASM failure rate of 24% is necessary, and that also
results in an unacceptable unproductive failure rate of 6.4%. This scenario
is illustrated in Figures 5.3.2 and 5.3.3.
The last scenario compares the tests at the maximum IDRs achievable while
limiting the unproductive failure rate to less than five percent. Again, the
IM240 IDRs are significantly higher than the ASM's, with a lower unproductive
failure rate. The IM240 HC IDR is 14% higher, the CO IDR is 14.3% higher, and
the NOx IDR is 17.5% higher, with an Ce rate that is 1% lower. Expressed
differently, an aggressive IM240-based program with a 3.5% unproductive
failure rate would relinquish about 14% of its HC and 17.5% of its NOx
effectiveness by substituting the ASM test at at an even higher unproductive
failure rate. This scenario is illustrated in Figure 5.3.4.
These statistics indicate that the ASM test is significantly less
effective than the IM240 as an I/M test.
The second scenario, wherein the ASM's HC IDR is raised to match the
IM240's HC IOR of 92%, is anticipated to raise the following question:
Why dkta't EPA neks the ASM*s HC onpoim mom stringent to increase
the ASM's IDR wfatant toatntaf the arafency of the ASKTi NOx aupoini.
The answe»-i« that eight vehicles (see Table 5.3.2) have a major impact on
the ASM HC XDR, but their ASM HC scores are less than 0.3 g/«i. Although
their ASM HC score* are very low, they account for roughly 10.5% of the total
excess FTP HC emissions. These eight vehicles also have ASM CO score* below
8.0 g/sd. While developing the ASM outpoint tables, EPA found that ASM
outpoints below 0.3/1.0 caused failure rates and Ccs to increase excessively,
so the final cutpoint tables did not include tighter outpoints. So to achieve
33
-------
the IM240'3 HC IDR, the only "practical" way to identify these cars is through
the NOx outpoint.
Five of the eight cars with high FTP HC that pass the ASM HC outpoints are
failed by a NOx outpoint of 1.5 or less. These five cars account for 7.6% of
the total excess HC emissions. So the 1.0 g/mi ASM NOx outpoint achieves an
HC IDR comparable to the 92.2% achieved by the IM240 at EPA'3 standard
outpoints.
Table 5.3.3 summarizes the ASM outpoint table in Appendix E to show that
the only way for the ASM to achieve the IM240's HC IDR of 92.2% at the
recommended outpoints for biennial programs is to lower the ASM NOx outpoint
to 1.0.
Table 5.3.2: Vehicle* that Paaa a
0.30 g/ai ASM HC Outpoint While Failing FTP HC
Vehicle HC FCT CO FTP NOx FTP HC ASM CO ASM NOx ASM
3180
3192
3195
3199
3201
3254
3257
3259
Table
ASM
0.96
0.49
0.51
0.53
0.94
1.87
1.26
1.94
5.3.
Cutpolats
0.30 / 8.0 t
0.30 / 8.0 >
0.30 / 8.0 /
0.30 / 8.0 1
0.30 / 8.0 /
0.30 / 8.0 /
0.30 / 8.0 /
0.40 / 8.0 /
1.00 / 8.0 /
1.00 / 9.0 /
1.00 / 8.0
' 2.0
1«*
' 1.9
1.4
1.3
1.2
1.0
1.0
1.2
1.0
/ 1.0
9.75
6.31
5.80
10.90
19.73
35.87
8.57
14.95
3: Alterni
Idoatifkatloa
HC 1 CO
88.0% 69.2%
88.2% 69.3%
89.0% 71.0%
90.3% 73.1%
90.3% 73.2%
91.4% 76.4%
96.6% 82.1%
92.4% 78.8%
84.1% 71.9%
91.3% 74.9%
92.4% 78.1%
1.22
0.53
0.66
1.53
1.72
1.16
0.90
0.53
0.15
0.20
0.18
0.29
0.17
0.29
0.25
0.23
itiva ASM Outpoints
••tea
1 NOx
74.9%
78.2%
82.7%
89.3%
89.3%
89.9%
93.0%
93.0%
89.8%
93.0%
93.0%
Fallart
Rat*
29%
30%
33%
38%
40%
42%
48%
43%
32%
40%
42%
Be
Rat**
4.3%
4.3%
4.3%
6.4%
6.4%
6.4%
8.7%
8.7%
4.0%
8.4%
8.4%
3.64
6.42
3.26
3.89
4.73
7.37
4.65
4.61
For High
I Dlscrtpaat
1 FallarM
0
0
42
42
84
12C
132
138
132
221
180
0.68
0.92
1.27
1.53
1.22
1.13
1.03
1.22
HC XDR4
1 UF
1 Rat*
4.3%
4.3%
6.4%
8.4%
10.4%
12.4%
13.0%
13.3%
10.4%
19.1%
17.1%
34
-------
To achieve an HC ZDR rat* greater than 89% a NOx outpoint of less than 1.5
is necessary. To achieve an HC ZDR rate greater than 91.4% a NOx outpoint of
less than 1.2 is necessary, and to achieve an HC IDR rate greater than 92% a
NOx outpoint of 1.0 is necessary. Once a tight NOx outpoint is used to fail
these cars with excess HC, the HC outpoint no longer determines the result, at
least in this sample. So, ASM outpoints of 1.00/8.0/1.0 are the least
stringent ASM outpoints that can achieve a 92% HC ZDR.
Another consideration is that the ASM outpoints have been optimized for
this database. Zn contrast, the ZM240 recommended outpoints were optimized
for the Indiana ZM240 database. Because of sample to sample differences, the
optimum outpoints are expected to vary slightly from one database to another.
So the optimum ASM outpoints are being compared to standard ZM240 outpoints,
which while optimum for the Indiana data, are not the optimum outpoints for
this database. Applying ASM outpoints optimized for this data base, to a
different database, is expected to further lower the ASM's performance.
35
-------
Figure 5 J.I
Comparison of ASM to IM240
Using IM240 Standard Outpoints
With Maximum ASM IDRs at Equivalent Failure Rates
100% -i-
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
83%
N
-------
Figure 5.3.3
Comparison of ASM to IM240
With ASM CO & NOx IDRs Matching IM240 IDRs
89%
COIDR NOxIDR
IIM24O«RecommeodcdQflpointi O ASM it Mttchfcf CO * NOx IDRs
38
-------
Figure 5.3.4
Comparison of ASM & IM240
Maximum IDRs <§> EC <5%
97%
COIDR NOxIDR FtfloreRuo UiyuJuuiw
• lM240Bc«IDIU9<3%Ec D ASM B«« HMU « <5% Be
39
-------
5.4 Comparison using Scatter Plots and Regression Tables
The objective of this analysis was to check the correlation of both the
IM240 and the ASM test with the FTP. The correlations are illustrated in
Figures F-l through F-9, in Appendix F. Appendix F includes regression tables
along with these scatterplots. The regressions show similar R2 over the
entire data range, but the ZM240 correlates much better to the FTP for
vehicles emitting closer to the FTP standards.
5.4.1 Using the Coefficient of Determination (R2 ) and
Standard Brror of the Estimate for Objective
Comparisons
R2 represents the percentage of variability in the dependent variable
(FTP result) that is explained by the independent variable (I/M test result)
and is often used to compare one I/M teat's effectiveness with another's, but
R2 can often be misleading. Since R2 is often used in correlation studies,
it does provide an indication of comparative test accuracy that would be of
interest to readers accustomed to seeing such comparisons. More important,
however, is how well these tests discriminate between malfunctioning and
properly functioning vehicles at an I/M station, which is best measured using
the techniques discussed in Section 5.2.
For a vehicle to fail an XM240, it must fail the two-ways-to-pass-criteria
developed by EPA (see Section 5.2.2). The R2 values presented in this section
are for composite IM240 scores only and do not account for this. Two-ways-to-
pass affects the quantitative correlation between IM240 and FTP significantly
because the Mode-2 HC and CO values are often more representative of vehicles'
actual FTP emissions. However, EPA believes it is not appropriate to mix and
match composite and Mode-2 scores into one quantitative correlation analysis.
Additionally, the R2 comparisons presented here do not account for the
sample's bias toward high emitters (discussed in Section 5.2.5.8). The 106
vehicles that were recruited to the lab for FTP testing were purposely biased
to include a high number, of dirty vehicles, when regressing the I/M test
scores verso* the) CTP to determine R2 values, these high emission values
disproportionately influence- some regression statistics, given the typical
distribution of in-use emissions data. Thus the emission values close to the
FTP standards) (where comparing I/M tests is most important), have less
influence on the R2 statistic than desirable for determining the actual
merits of these tests. Cutpoint tables account for this sample bias by
weighting each vehicles' emissions according to the population of vehicles
tested at the I/M lane.
40
-------
To account for these limitations the sample was divided into the following
three groups:
• All Vehicle*. This database is not very useful for comparing
correlation because the cleanest and dirtiest vehicles dominate the
R2 statistic. Both tests correctly differentiate these. More
pertinent are the vehicles with emissions closer to the FTP standard,
where the capability of I/M tests is not masked by the very clean and
very dirty vehicles. Also, vehicle 3211 is a CO outlier for both
tests. It has a major effect on the regression equation and the R2 ,
thus masking the typical capability for both procedures.
• Vehicle 3211 Removed for HC, CO, NOx. This database better
characterises the correlation of both short tests with the FTP, but
for the reasons discussed, it is not the most relevant for comparing
the effectiveness of the tests.
• Marginal Emitter*: Only vehicles that are not very clean or not
very dirty on FTP using following criteria:
HC 20.30 g/mi and <1.5 g/mi on the FTP
CO 22.5 g/mi and <25.0 g/mi on the FTP
NOx 20.5 g/mi and <2.25 g/mi on the FTP
Also, Vehicle 3211 was excluded as an outlier.
All vehicle* with FTP emissions less than 0.30 HC, 2.5 CO, and 0.5
NOx passed the ASM and IM240 tests, for all the cutpoint sets
evaluated in Section 5.3.
The standard error is an objective measurement of test variability
expressed in the unit* (g/mi. in thi* case) of the variable* used in the
regression. Because R2 are expressed as percents, standard errors have an
advantage of being leaa abstract.
Table 5.4.1 provide* a summary of R2 and standard error* for Figures F-l
through r-9 in- jppenrtim F, divided into the 3 group* just discussed. The
"Marginal Emitters* group indicate* that the R2 for the IM240 are
considerably higher for HC and NOx, and somewhat higher for CO. Likewise, all
the standard error* are lower for the ZM240, most notably for HC.
41
-------
Tabla 5.4.1 Summary of R2 and Standard Errors
Data Sat :
n :
Proeadura:
R* .
BC
Std. Error •
*' -
CO
Std. Error •
R* -
NOx
Std. Error •
All Vehicles
106
IM240
82%
0.62
54%
13.4
70%
0.65
ASM
73%
0.76
68%
11.2
71%
0.64
Vahicla 3211
Removed
105
IM240
83%
0.61
75%
10.0
70%
0.66
ASM
74%
0.75
80%
8.9
71%
0.64
Vehicles Naar
Standards
43
IM240
63%
0.19
25%
4.3
46%
0.34
ASM
17%
0.28
13%
4.6
26%
0.39
Tha standard arrors liatad in Tabla 5.4.1 can ba usad to astinata tha
lowast FTP value that would confidantly pradlct a dirty vahicla. For example,
tha HC standard arror La 0.28 g/mi for tha "Marginal Emitters" group. Since
95% of th« tin*, a vehicle's result will be within ±2 standard errors, this
suggests that tha lowest ASM HC score that confidently predicts an HC-dlrty
vehicle (i.e., FTP HC > 0.41) ia the ASM HC score that yields (using the
regression equation) an FTP HC of 0.97 g/mi [0.41 + (2 * 0.28)1. In contrast,
using the IM240 error of 0.19 g/mi means tha loweat IM240 score that
confidently predicts a HC-dirty vehicle is 0.79 g/mi, over 18% lass than the
score neaded to confidently predict an ASM HC dirty vehicle.
5.4.2 Advantage of Using Weighted Mode*
The ASM t«a% ia given a big advantage in the way the regressions are
performed baeaoa* each mod* is weighted separately according to the XM240.
the other hand, tn« XM240 score is a non-weighted scon. EPA developed the
ZM240 to contain similar driving conditions as the FTP. However, the
frequency of each condition ia not proportional to the FTP. By weighting
different modes of the ZM240 to the FTP similar to the way EPA has weighted
the 4 modes of the ASM test, EPA has found the R2 for the ZM240 to improve
immensely. The current score reported for the IM240 is something like
On
42
-------
weighting each mode of the ASM test 25%. This would hurt the correlation of
the ASM with the FTP, because, as is shown in Section 5.5, the 50 mph mode
accounts for roughly half of the composite ASM scores for each pollutant.
5.4.3 Observations of Scatterplots
The scatterplots for the first two sets of data (Figures F-l through F-6)
do not appear much different for either test, mainly because the high emittera
cause the emissions close to the standards to appear as a tight pack of data.
The plots for vehicles near the standard only (Figures F-7 through F-9),
however, suggest the following:
HC - The XM240 identifies the dirtier cars much better. Notice on the ASM
HC plot how many high emitters (FTP HO0.82 g/mi) still score
relatively low on the composite ASM score. Six vehicles pass the
very tight ASM HC cutpoint of 0.3 predicted g/mi, yet have FTP
emissions greater than twice the FTP standard (0.82 g/mi).
CO - Neither test appears to correlate very well over this emission range
for CO. Two issues come into play that explain why this is. First,
cars with loaded canisters will have high IM240 Mode-1 CO emissions
at the lane, causing the short test to have a high score while the
FTP at the lab is relatively low. The second scenario is cold start
problems. Two vehicle* in the database (Vehicles 3175 and 3227)
appear to have cold start problems, with high Mode-1 FTP CO
emissions, and low Bag-3 FTP CO emissions. Since the lane test is a
hot start test, these vehicles will show up clean at the lane, and
the cold start FTPS will be significantly dirtier.
NOz - The IM240 has a slightly tighter fit to the regression line, and more
of the FTP dirty care fall to the upper right of the scatterplot
(i.e., fail the teat properly).
5.4.4 Poosr- ASM WO Correlation
AS discussed in Section 5.3, ASM HC scores do not correlate very well with
FTP HC scores). Thia section briefly discusses theoretically why some of these
vehicles had very low ASM HC emissions, yet failed the XM240 and FTP for HC
emissions. Because the contractor'a mechanics were not aware of the ASM
scores, vehicles were not diagnosed with the objective of determining the
cause of the performance differences on these I/M testa.
The first four vehicles in Table 5.4.2 were found to have ignition
problems. Thia ia logical considering that misfire, which causes high HC, is
43
-------
sometimes related to load. A3 load increases the voltage required to jump the
spark plug gap alao increases. Some portions of the IM240 load the vehicle
more heavily than any of the ASM modes, so a marginal ignition system that
only causes misfire during the higher IM240 loads will not be identified by
ASM HC.
vehicles 3180 and 3264 were found to have bad 02 sensors and other
malfunctions. Slow responding O2 sensors are more likely to be identified
during a transient test, because changing throttle position tends to cause
air/fuel ratio excursions that will cause high emissions unless the fuel
induction system rapidly compensates to maintain the optimum air/fuel ratio.
So slow responding 02 sensor might explain the high HC on the IM240 and low HC
on the ASM.
The disconnected vacuum lines on vehicle 3201 could have caused lean-
misfire during accelerations on the IM240 that would not be apparent on the
steady-state modes of the ASM.
These explanations can not be proven with the existing data, but they
should indicate that a steady-state test suffers from known disadvantages
identifying vehicles with these types of malfunctions.
Table 3.4.2 Vehicles with Poor ASM HC Correlation
in
VIH ASM HC
HC XM240 HC Problem round
3259
3257
3155
3210
0.23
0.25
0.34
0.35
1.94
1.26
.25
.40
3180
3264
3254
3201
3165
0.15
0.49
0.2§
0.17^
0.3f '-
0.9*
1.3C
1.87
0.94
1.96
1.50 Ignition Module
1.92 Plug Mires, Plugs Transducer,
Ignition Coil Transistor
2.77 Incorrect Plugs* Torn wire boot
1.04 O2 Sensor, Spark Plugs, fuel Hose,
Catalyst
1.33 02 Sensor and Injectors
2.1C 02 Sensor, Vacuum Switching Valve
2.2S ECU Intermittent, Catalyst
1.15 Vacuum Lines Disconnected
1.59 Dirty Injectors
44
-------
5.5 Derivation of ASM Coefficients and Nod* Contribution
Variation* From Sample to Sample
This section discuss** why th« ASM coefficients that EPA basad its
analyses on war* developed using the IM240 as the dependent variable rather
than the FTP. BAR/Sierra and ARCO used the FTP to develop ASM Coefficients.
Also discussed are the rather large variations in the ASM coefficients
with different samples, and the variation in the contribution of each ASM mode
to the final ASM score (expressed as percent contribution).
5.5.1 ASM Versus ZM240 A* The Dependent Variable For
Determining ASM Coefficient*
EPA faced a dilemma in determining the best method for developing the ASM
equation coefficient*. No ASM advocate has recommended specific coefficients,
on which, EPA should accept or reject the ASM approach. So two options were
to: 1) perform the multiple regressions on all the lab recruited vehicles for
ASM versus FTP. Or, 2) perform the multiple regressions for ASM versus IM240
on a subset of the lane sampler excluding all lab recruited cars.
Obviously, the ideal method is to regress the ASMS versus FTPS (i.e.,
option 1), but this raises * problem. To evaluate how w*ll the ASM identifies
FTP failing vehicle*, the coefficient* must be applied to the lab recruited
vehicle* to calculate simulated grams/mile scores. However, good statistical
practice mandate* applying the coefficients to a different sample than those
from which they were developed.
This interlinking method, wherein the coefficients are applied to the sane
vehicle* from which they were developed, would minimise the effect* of the
test's variability. This improper interlinking is illustrated using results
from Vehicle 3211. Thi* vehicle'* lane-XM240 CO score wa* 93 g/mi and it* ASM
CO score wa* 65 g/mi u*ing the coefficient* from the 60S vehicle sample listed
in Table 5.5.1 (The- relevance of the other sample* in thi* table will be
discussed latex!. It* FW CO score was only 10.8 g/mi.
45
-------
Table 5.5.1
CO Coefficients Developed from Different
lea
Mode
Constant
5015
2525
50MPH
Idle
Adjusted R*
CO Sample 1
2.814
0.035
0.072
0.425
0.891
34.2%
CO Sample 2
2.836
0.116
-0.058
0.391
2.014
58.4%
CO All 608
2.936
0.040
0.043
0.356
1.350
50.1%
COvsFTP
5.533
-0.047
0.565
0.050
1.968
80.6%
The scatter plots below show that using the IM240-developed coefficients
cause this vehicle to be easily identified as an outlier. In marked contrast,
using the FTP-developed coefficients make it look like this vehicle's ASM
score highly correlates to its FTP score. The ASM mode scores are weighted
differently, so the high scoring mode(s) are de-emphasized. But these same-
sample FTP-based ASM coefficients are obviously peculiar to this sample, and
highly dependent on it containing this one particular car. (See Tables S.S.I
and 5.5.6.)
Still not answered is which ASM coefficients better indicate whether the
vehicle is malfunctioning or not. Some could argue that this vehicle should
not be an outlier. Instead, the XM240-developed coefficients inappropriately
make it appear as an outlier. Attempting to resolve this, the raw ASM
concentration measurements were checked. This vehicle's SO mph mode CO
concentration was 4.96%, which is higher than 97% of vehicles recruited to the
lab (103 of the 106 vehicles).
Using the same-sample, FTP-based ASM coefficients prevents this vehicle
from being an outlier because they adjust themselves to minimize the effect of
the SO mph mode score from all cars. Additionally, the very high concentration
measurement (4.96%) proves that the vehicle had a malfunction causing very
high emissions that had been inappropriately minimized. (This was also
verified during the mechanic's inspection which found a defective O2 sensor
and that an KM PROM update was required.) This evidence strongly supports
EPA's properly using the preconditioned ZM240s as the dependent variable for
developing AST coefficients to compare the ASM and XM240 correlations with the
FTP.
This evidence also casts doubt on conclusions developed from test programs
that used interlinked coefficients, interlinking makes the correlation
between the ASM, or any other test, and the FTP significantly better than
could be expected in an official I/M program. Since I/M programs will apply
ASM coefficients to vehicles that were never FTP tested, the opportunity for
46
-------
interlinking will not exist, so ASM performance should not be evaluated using
interlinked coefficients.
Another reason for EPA's not using FTP-based coefficients is because some
are negative, which means that as ASM501S emissions increase, FTP emissions
decrease. This is counter-intuitive.
47
-------
ASM Coefficients Developed
from
608 Lane ASM vs Pre-Conditioned IM240s
ASM Coefficients Developed
from
106 Lane ASM vs Lab FTP
8
48
-------
EPA decided that the best method, using the Arizona data, was to regress
the four steady-state nodes (three for NOx) against lane-only, preconditioned
!M240s. There were three major factors leading to this decision. First, this
allowed applying the coefficients to a different subset of data (the lab
recruited, vehicles). Second, the sample size was considerably larger (608 vs.
106 tests). EPA's FTP sample was too snail to divide and use half to
determine coefficients and the other half to evaluate ASM effectiveness, which
is supported by the negative coefficient yielded for the ASMS015 listed in
Table 5.5.1. Third, only preconditioned. IM240s were used because they
correlate better with the FTP than non-preconditioned. IM240s. The only
significant compromises in using XM240s instead of FTPS is that the composite
ASM score does not include a cold start excess (which would be independent of
warmed-up ASM node concentrations anyway) and that the mix of speeds and loads
in ZM240 is not exactly like that in the full FTP driving cycle (a hardship
borne by the IM240 in its own correlation to the FTP).
Figure 5.5.1 illustrates that preconditioned !M240s strongly correlate
with the FTP. These data are from the 106 lab recruited vehicles, but are
restricted to IM240s that were performed following the ASM at the lane, making
them preconditioned IM240s.
NOx has the worst correlation because of a few outliers at the high end,
but this ia not a concern for the ASM since the NOx coefficients are
relatively stable, which is discussed in the next section.
49
-------
M. JLJL1 Hicrli Correlation of Preconditioned Lan«-TM240a with FTPs
IM240vsFTP
Preconditioned IM240s
14
1.5 10 15 3.0
IM240HC
3.5 4.0
4.5
XM240vsFTP
Preconditioned IM240s
10
20
30 40 50 60
IM240CO
IM240vsFTP
Preconditioned IM240i
70 80
UO
24
3.0
IM240NOz
4jO
54
64
50
-------
3.3.2 Variability of ASM Coefficients
The objective of the following analysis was to investigate the stability
of the coefficients used to calculate composite ASM simulated grants/mile
scores. The database was divided into four samples for comparison.
Sample 1 was developed by using a random number generator to select 304
vehicles from the lane-only fleet of 608. The remaining 304 vehicles became
Sample 2. Sample 3 waa all 608 vehicles, and the fourth sample was the 106
laboratory vehicles. The ASM coefficients for the first three samples were
developed using the IM240 as the dependent variable and the FTP sample used
the FTP for the dependent variable. The resulting coefficients are listed in
the following tables. (Table 3.5.3 is a duplicate of Table 5.5.1.).
Table 3.5.2
HC Coefficients Developed froa Different Saaples
Mode
Constant
5015
2525
50MPH
Idle
Adjusted R*
HC Sample 1
0.080
0.045
0.047
0.147
0.084
21.7%
HC Sample 2
0.073
0.008
0.059
0.123
0.585
38.9%
HC All 608
0.083
0.025
0.059
0.136
0.124
29.0%
HCvsFTP
0.291
-0.261
0.507
0.238
0.154
79.4%
CO
Table 3.3.3
Coefficients Developed froa
Different Samples
Mode
Constant
5015
2525
50MPH
Idle
Adjusted R»
CO Sample 1
1814
0.035
0.072
*• 0.425
0.891
34.2%
CO Sample 2
1836
0.116
•0.058
0391
2,014
58.4%
CO All 608
2.936
0.040
0.043
0.356
1.350
50.1%
COvsFTP
5.533
•0.047
0.565
0.050
1.968
80.6%
51
-------
Table 5.5.4
NO* Coefficients Developed £ro» Different Sample*
Mode
Constant
5015
2525
50MPH
Adjusted Rz
NOX Sample 1
0.230
0.088
0.206
0.386
602%
NOX Sample 2
0.279
0.045
0.212
0.333
57.9%
NOX All 608
0.258
0.061
0.219
0.352
59.1%
NOX vs FTP
0.190
0.148
0.093
0.291
71.1%
The negative coefficients are highlighted in bold. One could infer from
the negative coefficients that increasing the emissions during that mode of
the ASM would lower the composite score.
These coefficients were used with the ASM data, froa each of the 106 lab-
recruited vehicles, to calculate the emissions for each mode and the percent
of the total emissions that each mode contributed. These mode contributions
give a better indication of each modes importance in the final ASM score, than
the coefficients, which are more difficult to interpret. The results are
listed in the following tables:
fable 5.5.5
Average) Contribution of Total HC
•sioos by Mod*
Mode
Constant
5015
2525
50MPH
Idle
HC Sample 1
17%
20%
17%
43%
2%
HC Sample 2
17%
4%
22%
39%
18%
HC All 608
19%
12%
22%
43%
4%
HCviFTP
11%
-75%
116%
45%
3%
52
-------
Table 5.5.6
Average Contribution of Total CO emissions by Mod*
Mode
Constant
5015
2525
50MPH
Idle
CO Sample 1
26%
4%
7%
57%
6%
CO Sample 2
27%
12%
-6%
53%
15%
CO All 608
29%
4%
4%
51%
11%
COvsFTP
28%
-5%
55%
7%
15%
Table 5.5.7
Average Contribution of Total HOx emissions by Mod*
Mode
Constant
5015
2525
50MPH
NOX Sample 1
15%
11%
24%
50%
NOX Sample 2
20%
6%
27%
47%
NOX All 608
17%
8%
27%
48%
NOX vs FTP
20%
22%
13%
45%
The HC coefficients in particular are very volatile, and that the negative
FTP-developed coefficients are counter-intuitive. When applying the
coefficients from Sample 1, the idle mode, on average, only contributes 2% to
the total score. This contribution jumps to 18% when the coefficients from
Sample 2 are applied. Similarly the ASM5015 contribution drops from 20% to
4%. These examples indicate that the largest sample (€08 vehicles) with
preconditioned ZM240s was the best sample available for developing ASM
coefficients.
5.5.3 Significance of Mode Contributions
The ASM mod* contributions also vary as the composite ASM score moves from
low values (fos) which the constant term will be the primary contributor to the
composite sce*e> to relatively high vslues (for which the constant term will
be a relatively small contributor to the composite emission). This is
illustrated in Figure* 5.5.2 to 5.5.4. For CO, th* ASM501S and ASM2S2S ar*
combined, because of th* negative contributions of 2525 and the small
contribution of the ASM5015 in relation to the 50 mph mode.
53
-------
rigur* 5.5.2
Hod* Contribution* for BC
U
staf Sample I Coeffiden*
I Constant Hldk QSOMPH 12525
5015
FTPHC4
FTP HC >
Ifofaf SMplt 2 Cotffidtnto
ldb DSOMPH B2523
5015
FTP HC > \M
54
-------
rigor* 5.S.3
Mod* Contribution* for CO
Using Sample 1 Coefficients
I Constant
DSOMPH
I 5015+2525
FTP CO S fcM
tM < CO S 13.M
134 < CO i 15*
Usfaf Sanple 2 Coefficients
lldte
DSOMPH
I 5015+2325
FTP CO *
55
-------
rigux* 5.5.4
Mod* Contribution* for HOx
100%
80%
60%
40%
20%
0%
Using Sample 1 Coefficfentt
! Constant DSOMPH B2525
5015
-I-
LJ
.lS 1.15
-------
The fact that the 50 nph mod* contributes so much to the composit* score
for each pollutant is also reason for concern. This opens the opportunity for
mechanics to adjust vehicles to lower emissions for just one mode (namely the
50 mph), which will be further discussed in Sections 5.6.3 and 5.6.4.
5.5.4 Conclusions on ASM Mode Contributions
While not a mode that was recommended by the ASM developers, the 50 mph
mode at road load horsepower appears to be more important for identifying
dirty vehicles than the lower speed, acceleration simulation modes (ASM5015
and ASM2525). Surprising was the small contribution of the ASM5015 (mode 1)
for identifying dirty vehicles considering that BAR/Sierra and ARCO both found
this mode to be the moat effective. This suggests that the first mode in a
four mode sequence serves mainly to precondition vehicles for the following
modes. Randomizing the order of the modes may be useful in determining the
best sequence.
For the outpoint analysis in Section 5.3 and the regression analysis in
Section 5.4, the ASM scores used were those calculated from the coefficients
developed from the 608 ASM versus preconditioned lane-ZM240s. However, the
variability of the HC coefficients between the two random subsets of 304 tests
suggest that a different sample of 608 tests might produce substantially
different equation coefficients. The resulting change in HC (and in some
cases CO and NQx) ASM scores would produce different failure rates, ZDRs, and
Be rates in the outpoint tables, and different R2 values in the regressions
of ASM versus FTP. So the volatile coefficients may vary from sample to
sample, or worse yet region to region, resulting in disparate X/M programs
which would be hard to evaluate on a consistent basis.
9.6 Repair Analyses
S.C.I Contractor Repairs
The objective) of this analysis was to investigate the performance of both
the XM240 antf the) ASM testa as predictors of changes (i.e., decreases or
increases) in FTP emissions following contractor-performed, IM240-target*d
repairs.
Of the IOC vehicles used in the outpoint analysis (Table 4.2.2), 56
exceeded the lane-IM240 0.80/15.0/2.0 + 0.50/12.0 outpoint and were repaired
by the contractor. Of these, 52 received each of the three following tests
both prior to repairs (i.e., as-received) and following repairs:
57
-------
- a Lane-IM240,
- a Lane-ASM, and
- an FTP.
These 52 were used in this analysis and are included with the data listed in
Appendix B. The resulting database of those 52 fuel-injected vehicles has the
following distribution:
Fuel
Metering
PFI
TBI
Order of
ASM Prior
to XM240
14
11
[•ana Testing
ZH240 Prior
to ASM
16
11
The contractor was instructed to perform the minimum repairs necessary in
order that each vehicle's IM240 emissions after repair (as tested at the
contractor's laboratory) meet the following criteria:
- composite ZM240 HC £ 0.80 g/mi,
- composite XM240 CO £ 15.00 g/mi, and
- composite ZM240 NOx S 2.0 g/mi.
The contractor was allowed multiple repair attempts if the first set of
repairs did not reduce the XK240 emission levels enough. The repairs were
limited to 91,000 per car. And, the contractor was instructed that "the
mechanic should only be aware of the XM240 scores for the XH240-targeted
repairs.* Because ASM outpoints, that could distinguish malfunctioning
vehicles from properly functioning vehicles, were not yet developed, only
IM240-targeted repairs were performed.
These IM240 emission repair criteria were met at the contractor's
laboratory foe all cars prior to the second and final FTP, with the highest
after-repair laboratory ZM240 composite HC emission score of 0.56, CO of
10.82, and WOm of 1.93 (g/mi). The effects of those IM240-targeted repairs on
FTP emissions are illustrated in the following table:
58
-------
«t 1 . 1
FTP Emissions Prior to and Following
IM240-Targeted Repairs
FTP Emissions
HC As -Received
After Repair
CO As-Received
After Repair
NOx As-Received
After Repair
Mean
1.458
0.326
19.707
3.331
1.649
0.739
Rang*
Xmisi
Minimum
0.16
0.10
0.28
0.63
0.20
0.05
• of
lions
Maximum
13.07
0.75
113.40
8.82
7.56
1.81
The resulting FTP emissions after the IM240-targeted repairs were essentially
independent of the as-received FTP emissions. (That is, the R-squares
associated with before and after HC, CO, and NOx were only 0.1%, 1.2%, and
1.0%, respectively.)
The data from these repaired vehicles can give insight into the question
of whether the IM240 test and outpoints cause repairs to be made which also
reduce FTP emissions. In other words, does the IM240 and the FTP correlate
well on a single vehicle? This correlation is to be expected based on the
realistic nature of the ZM240 driving cycle, and the good correlation found in
samples of vehicles not repaired.
For each of those 52 vehicles (all 1983 and newer fuel-injected cars), the
change in each pollutant (HC, CO, and NOx)• following contractor repairs, was
calculated for each of those) three test cycles. Regressing the reductions in
the lane emissions against the reductions in FTP emissions produced Tables
5.6.1.2 and &.6.I.3. The «ix graphs (Figures 5.6.1.1 through 5.6.1.3) that
follow those) regression tables illustrate the results of this analysis.
59
-------
S fi 1 g
Ragreaaion o* Changaa in Lana-XM240 Kaiaaiona relieving
Contractor Rapaira Varan* Corraaponding Changaa in FTP Bmiaaiona
Dependent variable
R2 .81.9%
is: A(FTP HC)
s- 0.8320 with 52-2.50 degrees of freedom
Source
Regression
Residual
Variable
Constant
AUM240 HO
Sum off Squares df
156.693 1
34.611 50
Coefficient «.e. of Coeff
-0.365173 0.1524
1.4106 0.0938
Mean Square
157
0.69222
t-ratio
-2.4
15
F-ratio
226
Dependent variable
R2 .47.5%
is: A(FTP CO)
s. 18.50 with 52-2-50 degrees of freedom
Source
Regression
Residual
Variable
Constant
AHM240 CO)
Sum of Squares df
15469.4 1
17110.1 50
Coefficient e.e. of Coeff
4.64057 3.103
0.846373 0.1259
Mean Square
15469
342.203
t-ratlo
1.5
6.72
F-ratlo
45.2
Dependent variable is: A(FTP NOx)
R2 .64.5%
s . 0.8846 wtth 52-2-50 degrees of freedom
Source -
Regression
Residual
Variable
Constant
AHM240 NOx)
Sum of Squares df
71.1008 1
39.1265 50
Coefficient s.e. of Coeff
-0.275563 0.1747
0.738523 0.0775
Mean Square P-ratio
71.1 90.9
0.78253
t-ratlo
-1.58
9.53
60
-------
R«gx*a«ioa ot Caang«« ia Lan«-ASM Kaissioaa rollowing
Contrmctor Repairs V«r«u« Corresponding Chang** ia FW Xalsaion*
Dependant variable
R2 -71.7%
is: A(FTP HC)
s- 1.040 with 52-2-50 degrees of freedom
Source
Regression
Residual
Variable
Constant
AfASM HO
Sum of Squares df
137.187 1
54.1169 50
Coefficient e.e. of Coeff
0.270944 0.1633
2.18967 0.1945
Mean Square
137
1.08234
t-ratlo
1.66
11.3
F-ratlo
127
Dependent variable
R2 -79.5%
is: A(FTP CO)
s- 11.55 with 52-2-50 degrees of freedom
Source
Regression
Residual
Variable
Constant
AfASM CO)
Sum of Square* df
25906.3 1
6673.29 50
Coefficient t.e. of Coeff
5.71775 1.775
1.0868S 0.078
Mean Square
25906
133.466
t-ratlo
3.22
13.9
F-ratlo
194
Dependent variable
R2 .70.8%
is: A(FTP NOx)
s - 0.8016). wtth 52-2-50 degrees of freedom
Source
Regression
Residual
Variable
Constant
A(ASM NO*)
Sum of Squares df
78.0956 1
32.1317 50
Coefficient e.e. of Coeff
-0.013624 0.1392
0.829714 0.0753
Mean Square
78.1
0.642635
t-ratlo
-0.098
11
P-ratlo
122
61
-------
rigor* 5.6.1.1
Decreases in HC Emissions Following Repairs
AFTPVSAIM240HC
Refresnoo Lin*
R-Squar«d«81J%
-1
135
AIM240 HC Emisrioos (a/ml)
AFTPTflAASMHC
AASMHCEBtatottfefel)
62
-------
3.6.1.2
Decreases in CO Emissions Following Repairs
AFTPvf AIM240CO
120
100
3
RcfradoB Line
R-Squarcd> 47.5%
. N
20 40 60
AIM240 CO Emtafou (g/mO
80
100
AFTPnAASMCO
20 40 6t
AASMCOBmhriow(i/»i)
100
63
-------
Figure 5.6.1.3
Oecreaaea in NOx Emissions Following Repairs
AFTPvsAIM240NOx
I 1
Rtfnttioo Line
R-Squ*rtd = 64.5%
135
AIM240 NOx Embriou (gfaQ
AFTFvsAASMNOs
1 3 S
A ASM NQx Eatarion (|M)
64
-------
Comparing the two graphs that examine the changes in HC emissions (Figure
5.6.1.1), it is apparent that one vehicle (vehicle number 3190) exhibited a
reduction in FTP HC emissions substantially greater than any of the other 51
vehicles (12.88 g/mi HC reduction compared to only 3.86 for the next larger
FTP HC reduction). Since it is possible that one such vehicle could
substantially affect the regression analysis, a second set of regressions were
performed on the remaining 51 cars (i.e., with vehicle number 3190 deleted) to
determine the effect. The effects on the slopes of the regression lines are
given in Tables 5.6.1.4 and 5.6.1.5.
Table 5.6.1.4
Kffeot oa ZM240 Regression Line
For HC Emissions
Of Deleting Vehicle 3190
Constant
Coefficient
R2
Bassd on Al
52 Vehicle*
-0.365173
1.4106
81.9%
Based on 51
Vehicle*
0.018227
0.93431
74.7%
Table) 5.6.1.5
Bffect oa ASM Regression Lin*
For HC Emissions
Of Deleting Vehicle 3190
52
on Al
Vehicles
Baaed on 51
Vehlclea
0.270944
I Constant
Ifft I 71.7%
0.452113
2.18967
1.35391
67.8%
From TaJble» 9.(.1.4 and 5.6.1.5, we see that deleting that potential HC
outlier (vehicle) number 3190) baa a similar effect on each regression line.
The slop* of the) ZM240 regression line decreases 11.6 degrees, and the slope
of the ASM regression lino decreases 11.9 degrees. Since delating the change
in HC emissions of vehicle 3190 from the sample has the same effect on both
regression lines, it would be) advisable to use the equations based on 51 cara
to estimate changee in FW HC emissions based on IM240 and/or ASM HC changes,
for IM240 and/or ASM HC changes between -1.0 and +4.0 g/mi.
65
-------
Comparing the two graphs that examine the changes in CO emissions (Figure
5.6.1.2), it appears, at first glance, that the composite IM240 tends to over
predict the repair benefit to CO emissions for some vehicles with relatively
small FTP CO repair benefits. However, the actual situation is that several
relatively cleaner vehicles (though still exceeding FTP standards) had
unusually high XM240 results on their first test. IM240 CO was a lot lower
after repair, but the FTP emissions had comparatively little room to improve.
The five vehicles in Figure 5.6.1.2 that exhibit this problem (vehicles
numbered: 3157, 3175, 3211, 3213, and 3214) all have as-received composite
FTP CO less than IS g/mi. For two of those five, most of the high composite
IM240 emissions resulted from the first mode (i.e., the first 93 seconds) of
the IM240. For this reason, EPA has recommended that vehicles which fail the
composite HC or CO outpoint be given a second chance to pass by examining the
Mode-2 emissions (see *Two-ways-to-paas" in Section 5.3). A similar situation
cannot happen for the ASMs as analyzed in this report because the weighting
factors, in effect, cause the CO scores on the first mode (5015) to be
ignored. One vehicle that deserves special note is vehicle number 3211. That
vehicle exhibited the largest ZM240 CO reduction (91.70 g/mi), but an FTP CO
reduction of only 8.00 g/mi. This high lane-IM240 CO reduction resulted from
a high initial (i.e., as-received) lane test score of 93.07 g/mi, but an
initial FTP CO score of 10.79. (However, the lane score was confirmed by an
indolene-fueled lab-IM240 following the FTP which had a CO result of 52.48
g/mi.) The ASM tests on this vehicle did not exhibit a large CO reduction
following repairs because bath the initial ASM and the ASM following repairs
exhibited very high CO emissions (more than 5%) during the 50 mph cruise mode.
(Thus, the ASM did not over estimate the CO repair benefit on vehicle 3211
because the ASM over estimated both the initial FTP CO emissions, as well as,
the FTP CO emissions following repair.) In apite of the few over predictions
of emission benefits from repairs, it should be noted (as illustrated in Table
5.6.1.1) that following the IM240-targeted repairs, no vehicle was left with
high unrepaired FTP emissions.
Most of the} vehicle*, which exhibited very little if any HC or CO
improvement following the) IM240-targeted repairs, had been recruited for
repairs becmae> they exhibited, on the lane-lM240 test, low HC and CO, but
high NOx- Therefore, no significant improvement in either FTP HC or CO was to
be expected.
Comparing the two graphs that examine the changes in HOx emissions (Figure
5.6.1.3), it is apparent that on* vehicle (vehicle number 3202) exhibited a
reduction in FTP NOx emission* greater than any of the other 51 vehicles (6.31
g/mi NOx reduction compared to 4.98 for the next larger FTP NOx reduction).
Since it is possible that one such vehicle could substantially affect the
regression analysis, a second set of regressions were performed on the
remaining 51 cars (i.e., with vehicle number 3202 deleted) to determine the
66
-------
effect. The effects on the slopes of the regression lines are given in Tables
5.6.1.6 and 5.6.1.7. From Tables 5.6.1.6 and 5.6.1.7, we see that deleting
that potential NOx outlier (vehicle number 3202) has virtually no effect on
either regression line. The slope of the XM240 regression line decreases only
3.3 degrees, and the slope of the ASM regression line decreases less than half
a degree.
Table 5.6.1.6
Effect oa XM240 Regression Lin*
For HOx Emissions
Of Deleting Vehicle 3202
Constant
Coefficient
R2
Based on Al
52 Vehicles
-0.275563
0.738523
64.5%
Based on 51
Vehicles
•0.183932
0.652265
57.4%
Table 5.6.1.7
Bffeot oa ASM Ragressioa Liae
rox MO* Emissions
Of Deleting Vehicle 3202
Constant
Coefflelen
R*
Based on Al
52 Vehicles
•0.013624
0.829714
70.8%
Based on 51
Vehicles
•0.003336
0.816328
60.1%
Six vehicles (vehicle numbers: 3172, 3200, 3212, 3239, 3240, and 3244)
exhibited large decrease* ia lane NOx emissions, but little if aay change in
FTP NOx •ttiaeioas). These six had a number of factors ia coonon:
All six had low as-received FT? HC (for five of the six HC £ 0.37,
and BC • 0.59 for the sixth), CO (CO £ 3.47), and NOx (NOx * 2.34) .
- All six had low as-received lane-IM240 HC (HC 1 0.29) aad CO
(CO £ 3.33), but high laae-XM240 NOx (NOx * 1.14).
67
-------
six had low as-received ASM composite HC (HC £ 0.24) and CO
(CO 0.80 and Mode-2 ZM240 HC > 0.50.
2) the 16 vehicle* that were recruited (and repaired) because their
initial lane-XM240 exceeded the CO cutpoint of:
Composite XM240 CO > 15.00 and Mode-2 XM240 CO > 12.00.
3) the 30 vehicle* that were recruited (and repaired) because their
initial lane-DC40 exceeded the NOx cutpoint of:
Composite IM240 NO* > 2.00.
As previously discussed, two vehicle* (vehicles numbered 3211 and 3190) could
be deleted from the -HC-Repaired" and from the "Co-Repaired" data bases due co
questionable test result*. Additionally, vehicle number 3202 could be deleted
from the "NOx-Repaired" data, base for similar reason*. Thus, in addition to
performing regression analyse* on the entire 52 car data base, we can also
perform regression* on the 32/16/30 (HC/CO/NOx) subset*, a* well a*, (after
deleting the questionable vehicle*) on the 30/14/29 car subset*. Within these
various data seta, we performed 1C linear regressions, the results of which
are summarized in Table* 5.6.1.8 through 5.6.1.10.
68
-------
^ g
R«gr«««ioa Lin«« of ABC for Short
V«r*ua FT*
Conatant
J^A^^fl^BlA^A
bovniCMni
R-Squaiad
....... ••••^»
Baaad on
MB
VaMclaa
-.365173
1.41060
81.9%
BaMd on
32
ExcMdlng
InftlalHC
-.932339
1.63036
82.7%
BaMd on
32lft«»
TWO
0.014741
0.958254
633*
BaMd on
AIS2
VaMClM
.270944
2.18967
71.7%
Baaad on
32
ExcMdlng
inMalHC
0.371352
2.14853
672%
Baaad on
32 Minus
Two
0.7929
1.11862
903%
a. a. i. a
R*gx«»ioa LiM« of AGO for Shore T«at« y«rao«
Conatant
CoafUdanf
R-Squavad
•••••••••
Aisa
VaMdM
4.64057
0. 841373
47.5%
— IM240-
«
Eicaartlnq
InttMCO
17.S097
0.611959
18.7%
BaMd on
lOlflnua
Two
-0.36712
1.28527
55.1%
~.~_. •«•*
Baaad on
Alfl2
VaMdM
5.71775
1.08685
79.5%
Baaad on
tt
Excaadlng
imunco
13.9744
0.95755
73.0%
Baaad on
l6Mnua
Two
13.4061
0.962594
74.1%
69
-------
5 £ I 1ft
Regression Lines of AMOx for Short Tests Versus FTP
Constant
Coefficient
R-Squarad
Baaadon
AI82
Vahtolaa
-.275563
0.738523
64.5%
___ IU94A _
Baaadon
a
Exceeding
Initial NOx
-.778908
0.886525
54.2%
BaMdon
On«
-0.45849
0.733666
39.9%
Baaadon
Aiaa
Vehicles
-0.013624
0.829714
70.8%
.... AQU ...
Baaadon
30
Exceeding
Inttlal NOx
0.209571
0.763686
62.3%
Baaad on
30 Minus
Ona
0.273927
0.711124
45.7%
Examining the slopes and y-intercepts (i.e., the "coefficient* and
"constants" in Tables 5.6.1.8 through 5.6.1.10) of the 18 regression lines, we
make the following observations:
- Limiting the analysis to only those vehicles whose initial lane-!M240
test exceeded the outpoint for the pollutant being examined:
— had virtually no effect on the regression line predicting FTP
HC changes based on ASM HC changes, and only a relatively small
effect on the line predicting FTP CO changes based on ASM CO
changes;
— had moderate effects on the two regression lines predicting FTP
HC and CO changes based on IM240 HC and CO changes; and
— had only relatively small effects on the regression lines
predicting FTF NOx changes based on ASM or XM240 NOx changes.
Again, the) effect was larger for the IM240 case.
- Deleting the one or two questionable vehicles prior to performing the
regression analysis:
— produced only small effects in the two NOx cases (IM240 and
ASM) and in the ASM CO case and
— produced substantial effects in the two HC cases and in the
IM240 CO case.
In summary, this analysis indicates that the change in ASM scores before
and after repairs correlates with changes in FTP emissions* about as well as
70
-------
for the IM240. However, because ASM outpoints were not recommended by ASM
proponents and EPA did not have outpoints to use as repair targets/ the
contractor repairs were performed to attain XM240 scores that complied with
the standard ZM240 outpoints. So the contractor repairs offer little insight
into the primary question of whether vehicles repaired to pass an ASM test
will be as effective ae vehicles repaired to pass the IM240 test. The next
two sections will further discuss repair issues.
5.C.2 Commercial repair*
5.6.2.1 Introduction
The purpose of this analysis was to compare the effects of commercial
repairs, for vehicles that failed the Arizona I/M test, on IM240 and ASM
after-repair test results. Experience has shown that commercial repairs
geared to steady-state I/M tests have not met expectations for in-use emission
reductions. Because vehicles are operated only at steady-state, repairs have
been geared to reducing emissions at those operating conditions. As a result,
emissions over the full range of operating conditions are often not
effectively reduced, even when vehicles are repaired to pas* a steady-state
I/M test. Thia is one reason EPA has established a transient test for
enhanced I/M. Since the XM240 requires vehicles to perform over a wide
variety of real-world operating conditions, IM240-successful repairs must be
effective in reducing emissions over a wide range of operating conditions.
By comparing the effects of commercial repairs on ASM and IM240 test
results at selected outpoints, an evaluation of the comparative repair
effectiveness can be made. As discussed above, EPA analyzed the results of
repairs performed to pass the Arizona I/M test to determine whether such
repairs would significantly reduce ASM emissions without significantly
reducing FTP emissions. Sine* the ASM test and the Arizona I/M test are
somewhat similar in that they are steady-state tests, repairs for the Arizona
I/M test may provide) information on whether ASM-successful repair are as
effective as> XM240-effactive repairs. The data show that successful repairs
for the Arizona) I/M teat are more likely to be successful for the ASM test
than for tha- IH240.
S.C.2.2 Database/Analysis
EPA1a commercial repair program in Mesa consisted of offering incentives
to owners of 1983 and newer vehicles that failed the Arizona I/M taat, but
were not needed or declined to participate in laboratory testing, to return to
EPA'3 IM240 lane for after-repair ASM and IM240 tests. To receive their
incentive, they were told to return with a receipt for commercial repairs. No
71
-------
instructions were given to owners regarding where or how to get their cars
repaired, and owners were not compensated for the actual repair itself.
As of April 1, 1993, before- and after-repair data were available for 23
of these vehicles. One vehicle, 113239 (CRf 24) was removed from the database
due to unacceptable speed deviations on its initial ASM test, leaving 22
vehicles available for analysis. For this analysis, five other vehicles were
excluded because they continued to fail the Arizona test after repairs. The
resulting database consisted of 17 successfully repaired, 1983 and newer
vehicles.
Cutpoints were applied to the IM240 and ASM data to determine pass/fail
status. The pass/fail determinations were then compared to evaluate the
effects of commercial repairs. Three different outpoint sets were used to
make the comparisons. Since the Arizona test measures HC/CO only, the first
comparisons involved only HC and CO criteria. Two additional comparisons were
made which included NOx cutpoints. All three are listed below (Section 5.3
discusses the relevance of these cutpoints.):
• IM240 recommended cutpoints for HC/CO with ASM cutpoints that produce
the highest IDRs at the same failure rate as the IM240 recommended
cutpoints:
IM240 - 0.80 / 15.0 -I- 0.50 / 12.0
ASM - 1.00 / 8.0
• EPA recommended XM240 cutpoints including NOz with ASM cutpoints that
produce the same 18% failure rate. These cutpoints are listed below:
ZM240 - 0.80 / 15.0 / 2.0 + 0.50 / 12.0
ASM - 1.00 / 8.0 / 2.0
• ASM and ZM24O outpoints selected to achieve the highest ZDRs possible
whil*> keeping th* probable EC rate below St. These cutpoints are
listed below:
IM2«r- 0.30 / 9.0/ 1.7 + 0.19 / 7.0
ASM - 0.40 / 8.0 / 1.5
For each set of cutpoints, a comparison of th* initial and final test
results were mad*. To *valuat* th* effects of repairs on a specific Z/M test
a vehicle must b* Identified by th* Z/M test for repair*. Thus, while the
initial test result comparison allowed th* identification ability of these two
X/M tests to b* compared, th* final test result allow* an evaluation of the
relative repair effectiveness of th* Z/M tests. Th* data were restricted to
72
-------
vehicles which were identified by all teats for the comparison of final test
results. Using these common vehicles allows the comparison of repair effects
to be clearly illustrated.
The results, which are discussed in the next section, indicate that the
IM240 is superior at identifying vehicles requiring repair and that for
vehicles which initially fail both the ZM240 and ASM, steady-state repairs ace
more likely to result in ASM passing scores than in XM240 passing scores.
5.6.2.3 Results/Conclusions
Initially, all 17 vehicles used in this analysis failed their initial
Arizona I/M test. However, the XM240 and ASM identified slightly different
sets of vehicles as needing repairs. Vehicles of interest are those that pass
the initial IM240 and fail the initial ASM and those that fail the initial
IM240 and pass the initial ASM.
As shown in Table 5.6.2-1, for the initial HC/CO only comparison, one car
passed the IM240 and failed the ASM and four cars passed the ASM and failed
the IM240 (see Appendix B for data listings). These errors-of-onission
support the assertion mad* in Section 5.3 that the ASM is weaker than the
IM240 at identifying malfunctioning vehicles with HC and/or CO emission
problems.
Table 5.6.2-1
Initial Pass/Fail Status Comparison
HC/CO only
Cutpointe
IM240 ASM
PASS PASS 3
IM240 ASM
FAIL PASS 4
IM240 ASM
PAS* PAIL 1
ZM246& ASM
nas FAIL 9
Comma Failure
Kate Cut points
ZM240 ASM
PASS PASS 2
XM240 ASM
FAIL PASS 2
ZM240 ASM
PASS FAIL 0
ZM240 ASM
FAIL FAIL 13
Optim*! XDR/MU
Be Cutpoiate
ZM240 ASM
PASS PASS 1
ZM240 ASM
FAIL PASS 2
IM240 ASM
PASS FAIL 0
ZM240 ASM
FAIL FAIL 14
Vehicle 13504 (CR* 25) failed the ASM and Arizona test duo to a CO problem
which appear* to occur only at idle operation. Because the ZM240 driving
cycle includes little idle operation, this vehicle was not identified by the
IM240 HC/CO only outpoints. An air/fuel mixture adjustment reduced emissions
sufficiently to pass the HC/CO outpoints for the ASM and Arizona teats.
However, this vehicle did exhibit excessive NOx emissions that were identified
by the addition of a NOx outpoint. Incidentally, the fuel mixture adjustment
4
73
-------
did little to address or reduce this vehicle's NOx emissions on either the
IM240 or the ASM.
In contrast to the IM240, which failed to identify only one vehicle, four
vehicles passed the ASM HC/CO only outpoints and failed the IM240 and Arizona
outpoints. Three of these vehicles are examples of ASM errors-of-omission and
illustrate the superior identification ability of the IM240. The fourth
vehicle failed NOx and will be discussed after the three that passed.
Vehicle 13471 (CRt 27) failed the Arizona and IM240 tests because of high
CO emissions, but passed the ASM test. Vehicle 13125 (CRt 12) failed HC on
both the IM240 and Arizona tests and was not identified by the ASM outpoints.
Vehicle 13202 (CRf 15) failed the HC and CO idle modes of the Arizona test.
On the IM240, vehicle 13202 failed HC and NOx but passed CO due to the two-
ways-to-pass algorithm. The ASM identified this vehicle for NOx emissions
only.
The fourth vehicle that initially passed only the ASM test was vehicle
12771 (CRt 8). This vehicle exemplifies the weakness of steady-state X/M
tests and is discussed in detail in Section 5.6.3. Vehicle 12771 failed CO
on the loaded mode of the Arizona test but passed the CO outpoint on both the
IM240 and the ASM. However, the car failed NOx and HC on the IM240 and failed
only NOx on the ASM. After repair, this car passed both the ASM and Arizona
tests even when ASM outpoints were tightened. These repairs did not
sufficiently reduce emissions over the full operating range of the vehicle,
demonstrated by the vehicle continuing to fail both HC (1.01 g/mi) and NOx
(3.01 g/mi) on the ZM240. This supports the assertion made in the
introduction that repairs to pase a steady-state test may not be effective in
reducing emissions over normal driving conditions and, therefore, do not
effectively reduce in-use emissions.
To illustrate the effects of commercial repairs on ASM and IM240 after-
repair test results, data were restricted to vehicles that failed both the
initial ASM and ZM240 (see Table 5.6.2-1). The results of these comparisons
are graphically depicted in rigures 5.6.2-1 thru 5.6.2-3.
74
-------
Figure 5.6.2-1
ircial Rapaira Paaaing ASM and IM240 Cutpointa
HC/CO only Coapariaon
Pass ASM
But Fail
IM240
Figura 3.6.2-2
•roial Rftpaica Vaaaiag ASM and ZM240 Catpointa
Rat* Coapariaon
P«S« ASM
But Fail
IM240
75
-------
Figure 5.6.2-3
rcial Repairs Paasing ASM mnd ZM240
High Ko Comparison
14
Cutpointa
Pass
Final
IM240
Pass ASM
But Fail
IM240
These graph* show that vehicles can and will be repaired to pasa the ASM
test but will continue to fail the IM240.
For the first comparison using only HC and CO outpoints, three vehicles
passad the ASM but continued to fail the IM240. The second comparison added
the NQx outpoint which in combination with the HC/CO outpoints produced the
same failure ratea for the XM240 and ASM. Again, three vehicles passed the
ASM but continued to fail the XM240. For the comparison using the most
stringent outpoints for the IM240 and ASM, five vehicles passed the ASM but
continued to fail the IM240. For all of these comparisons, there were no
vehicles that failed the ASM and passed the IM240 after commercial repairs.
This indicates that repairs which are sufficient to pass the ASM test are not
necessarily sufficient to pasa the IM240, indicating that the repair
effectiveness of the IM240 ia superior to that of the ASM.
Based on these results, repairs to pass the steady-state Arizona I/M test
are significantly more effective at reducing ASM emission scores than XM240
emission score*. Although the sample of successful commercial repairs is
small, these reaulta indicate that the ASM test, if implemented, will result
in significantly lower identification rates and emission reduction benefits
than those of the ZM240. A more detailed investigation of ASM emission
reduction benefits ia discussed in Section 5.6.3.
76
-------
3.6.3 Xa-Use emission Reductions from Steal World
Repairs
On* of the concerns with any t«st is the ability of an observed reduction
on the test to reflect real and permanent in-use reductions. Two particular
concerns are (1) can unscrupulous mechanics find repair strategies that would
allow a vehicle to temporarily pass the I/M test without resulting in
permanent in-use reductions (i.e., temporary repairs would be undone after
passing the test), and (2) is the test sufficiently imprecise such that
honest, but insufficient, repairs would not be detected by the I/M retest.
Teat Defeating Strategies
It is common knowledge that the current idle test can be, and is being,
defeated by a variety of methods. Most can be used only in the privacy of a
test-and-repair station. Some of the common ones that can be used in a test-
only station include creating a vacuum leak to lean out the air-fuel ratio for
CO failure*, and raising the idle speed to create a similar effect. A logical
question is, what ia the likelihood that test defeating strategies can be
developed by unscrupulous mechanics for the ASM or for the XM240 I/M tests.
On the surface, the ASM teat appears easier to beat than the ZM240 because
of ita steady-state nature and number of limited operating modes. In theory
at least, the mechanic could employ a similar method to the idle test for ASM
CO failurea. The proceaa would include creating a vacuum leak and disabling
the feedback control system. Since it is assumed that moat shops would have a
dynamometer in an ASM I/M scenario, the mechanic would simply need to operate
.the vehicle on the dynamometer and adjust the leak until the car was under the
outpoints. Moat likely the driveability of the car would be quite poor;
however, it would only need sufficient driveability to drive to the test
center and return, where the teat beating repairs could be undone.
If the vehicle could drive to the test center, then it could certainly
drive the steady-state teat, since driveability ia not required oa the ASM,
and emisaioae>•*• not recorded during the transitions between ASM teat modes.
Conversely* tme> emiaaione are measured during driving transitions on the
IM240, and the* lack of driveability would require more throttle movement with
a likely subataatial increeae ia CO emissions. If miafire occurred during
driving tranaitiona because of the leaa condition, the HC, and possibly NOx,
would increase oa the XM240, but would not on the ASM (because emissions are
not measured during driving transitions).
Another potential teat defeating atrategy that could occur oa the ASM for
NOx failurea deela with ignition timing. Retarding ignition timing has long
been aa approach to reducing HO*. Retarding the ignition timing excessively,
77
-------
however, reduces driveability. Once again, however, driveability is not
required on the ASM. A severe loss in driveability on the IH240 would be
expected to increase CO significantly, but would be expected to have little
effect on the ASM CO levels.
Some may point out that many new cars do not have adjustable distributors,
and others do not have distributors at all. Therefore, it would not be
possible to retard the timing, so such a test defeating strategy would not
exist on these cars. What many may not know is that all cars with non-
adjustable distributors and those without distributors have a base timing mode
that can be activated. Activation of base timing will severely retard the
timing in most cases, and could be used to lower NOx emissions.
Since these are only a few of the less creative methods that might be
attempted to defeat an ASM or XM240 test, it would be useful to verify if the
theoretical potential really could occur. Currently there is no data on
purposefully test defeating repairs. However, data from vehicles tested in
Arizona and sent for commercial repair may shed some light on the potential
for test defeating or improper repairs to be identified by either the ASM or
the XM240.
In our commercial repair data base, twenty-two vehicles failed the Arizona
X/M test which includes a steady-state loaded mode and an idle mode. All of
these vehicles received a 4-mode ASM test and an IM240 test. The vehicle
owners took the vehicles for commercial repair, and volunteered for repeat ASM
and XM240 tests when they returned for their Arizona retest.
Five) of the 22 vehicles were excluded from this analysis because their
four-mode ASM emissions did not exceed the outpoints of 1.0/8.0/2.0
(HC/CO/NOx). The resulting 17 vehicles represent the portion of the 22 car
sample that would have failed a four-mode ASM test if that had been the
official teat. Not* that this group of 17 vehicles represents a different
portion of the> sample) of 22 coonercially repaired vehicles than the 17
vehicles used for analysis in Section 5.6.2. The analysis in Section 5.6.2
excluded five* vehicle* that ultimately did not pass the Arizona X/M test after
repairs. TBe> analysis in this section excluded five vehicles that passed the
initial ASMVtaat, but included those vehicles that did not ultimately pass the
Arizona X/M test after repairs.
The repairs conducted on th« 17 vehicles are listsd in Table 5.6.3-1.
From these repairs and the) resulting ASM and IM240 scores* the possibility of
test defeating atrategies can be evaluated. Mote that the multiple repairs
represent retest failures on the Arizona X/M teat. Also, four of the 17
vehicles that initially failed the ASM test did not ultimately pass the
Arizona X/M teat.
78
-------
The before and after repair emission results are graphically represented
in Figures 5.6.3-1 though 5.6.3-3. From the repair data reported by the
commercial garages (in Table 5.6.3-1), it is clear that many vehicles had the
air-fuel ratio adjusted or received repairs that would likely affect air-fuel
ratio. Many of the vehicles were feedback carbureted; however, this should
make no difference for the purposes of evaluating the effect of air-fuel ratio
on the test type. Only vehicles CR-07 and CR-16 had reported commercial
repairs that would not likely affect air-fuel ratio (it was assumed that the
"tune-up" repairs in Table.5.6.3-1, in some cases, could have involved
adjustment of air-fuel ratio).
On these other vehicles, the degree of the effect oa air-fuel ratio is
unknown. But, from the CO emission results in Figure 5.6.3-2, it is clear
that in general, a repair that resulted in reduced CO on the XM240 also
reduced CO on the ASM. However, there are some exceptions. These are
vehicles CR-10, and CR-25. Vehicle CR-10 failed the before and after IM240,
failed the before-ASM, but passed the after-ASN. Whereas vehicle CR-25 passed
the before and after IM240, failed the before-ASM, and passed the after-ASM.
Since CO is primarily a function of air-fuel ratio* the observation from
these two vehicles ia that the air-fuel ratio during the steady-state test can
be different than the average over the transient test. To some extent, this
observation also appears to be evident in the CO results for vehicles CR-03,
CR-06, and CR-22 (see Figure 5.6.3-2). In the case of vehicle CR-10, the air-
fuel ratio during steady-state operation is sufficiently lean after repairs to
allow the vehicle to pass the ASM, but rich enough overall during transient
.driving to cause an IM240 failure. The opposite is apparently true for
vehicle CR-25, where the before repair air-fuel ratio during steady-state is
apparently sufficiently rich to cause an ASM failure, but lean enough during
average driving to allow the vehicle to pass the XM240.
Certainly, the> CO leTol can also be affected by the catalyst. But the
catalyst was) the> same ia all of these tests, so the catalyst effect should
wash out. Alaev catalyst efficiency can be somewhat gauged by HC levels as
seen in Figure* 5.€.3-1. The after repair HC levels oa vehicle CR-10 clearly
pass the ASftt eotpoiat. Thsj after repair IM240 HC status parallels the CO
status. la other words), based oa the ZM240 this vehicle was still broken, but
was passed oa the ASM. The HC levels oa vehicle CR-25 were low for all IM240
and ASM tests. Based oa ZM240 results, this vehicle should not have been
failed for HC or CO. However, vehicle CR-25 did have serious problems as
evidenced by the MOac emissions ia Figure 5.6.3-3.
The emissioa results oa these two vehicles, reinforce the following point.
Air-fuel ratio caa affect the CO levels on both tests. In particular, the
79
-------
s.
air-fuel ratio during a steady-state mod* can be different than the overall
ratio during transient operation. Therefore, it is likely that with willful
intent, a mechanic could purposefully create a vacuum leak, and adjust it so
that a car could pass the ASM, but not the XM240. Whereas the amount of
leanness in vehicle CR-10 was not sufficient to pass CO on the ZM240, it was
sufficient to pass the ASM. Furthermore, the amount of leanness was not
sufficient to cause vehicle CR-10 to fail either the IM240 or the ASM NOx
outpoints. Therefore, the results on vehicle CR-10 support the theoretical
possibility that unscrupulous mechanics could, with proper adjustment of
vacuum leaks, be able to adjust vehicles to temporarily pass the ASM CO
without increasing the NOx emissions sufficiently to cause an ASM NOx failure.
As indicated previously, the likelihood of such improper and temporary repairs
would be exacerbated in a program where the ASM was the official test, because
unscrupulous repair centers could conveniently maladjust a vehicle on a
dynamometer to pasa the steady-state modes of the ASM test.
80
-------
Tablt 54.3.1
Vehicle Repairs
VEttNQ 1ttPq»ir
CR-01 Adjusted aovTuel mixture on
CR-02
CR-03
CR-04
CR-05
CR-06
CR-07
CR-08
CR-09
CR-10
CR-13
CR-15
CR-16
CR-21
I ata/fuel mixture on
Adjusted aa/tod mixture on
r. Replaced heat valve.
vacuum leak and adjust
ignition timmf.
Bfptactxt O2 Sensor and performed
TuneUp.
Rpl O2
cap and rotor,
Rpl te bd^aa, fud flt AdJ
QfltflUL ^^IttDflBQ OlL
TUM« op, replaced fuel fltar,
^» 4na^^*
» nnBT.
AdJuswJ attnet nrixtnre and kite
AdjusttdaoTTud mixture, idte speed.
•fearer*
Set
Rpred electrical short in
harness from ECU to mix
control.
Adjusted Ukxaa/fud
mixture, cleaned fuel
injectors.
A one year waiver was
J^^M^»dl &.« *kl« -•-•-• —
giauiBu tor uiii vrnrMii
Adjusted ata/tal mixtnre
and kflA!
Performed Tdne-ap.
iper operation of choka Scoped <8
-------
-''«,
lUJ.t
Commercial Repair Effects
to MC briMtam « CM
MMHC.CO.WNOl
3 I I \ 1 I S3 \ I I I II I I
8
Commercial Repair Effects
I I I I I I II \ II I I 3 I I ?
VtMeto Hunter
82
-------
Commercial Rtpair Effects
33333333333353333
Insufficient Repaira
Another concern with an I/M teat ia the ability for the teat to cause
'proper and sufficient repaira to be performed if the vehicle faila the I/M
teat. For this analyaia, proper and aufficient repaira are conaidered to be
repaira sufficient to pass the IM240 outpointa. The commercial repair data
used in the preceding section on test defeating strategiea can alao provide
some inaight into thia iaaue.
Of internet* ia the) coapariaoa of teat modea between the 4-mode ASM and the
Arizona I/It tee*. Both have aa idle mode, and both have a steady-state loaded
mode. The. Adaona loaded mode is similar to the ASM 2S2S mode.
Using the general aisdlarity of the teat (i.e., idle and loaded modes),
the general sufficiency of ASM repaira can be approximated by observing the
results fro» vehiclea used ia the previous section that failed the initial ASM
teat and the initial Arixona I/M teat. A caae history oa vehicle number CR-
08, which initially failed the ZM240 HC and NOx outpoint* (aa well aa the
Arizona CO outpoint and the ASM MQx outpoint), illustrates the concern about
the ability of the ASM teat to cauae proper and sufficient repaira to occur
in-use.
83
-------
After vehicle number CR-08 had failed the Arizona Z/M test for CO on
January 4, and had received initial tests on the ZM240 and ASM, it was
enlisted in the commercial repair program. Two weeks after the commercial
repairs (January 14), the vehicle returned for its after-repair IM240, ASM,
and Arizona Z/M retest (and for the owner to obtain the recruitment incentive
payment). At that time, it was discovered that two days after the initial I/M
test (which was conducted on January 4), and following repairs (listed in
Table 5.6.3-1), the repair center had taken the vehicle to another Arizona
test lane for an Z/M retest. At this other z/M lane, on January 6 the vehicle
easily passed the Arizona outpoints of 1.2% CO and 220 ppm HC. However, when
retested on January 14 at the IM240 test lane, this vehicle failed the Arizona
HC outpoint by a wide margin (see Table 3.6.3-2). The owner was demonstrably
upset (even though a valid Arizona passing certificate had been issued), and
left the teat center abruptly. However, the owner returned again in another
two weeks (January 26). At this time, the vehicle passed all of the Arizona
outpoints. The owner did not divulge any information on corrections or
repairs that may have occurred between January 14 and January 26.
Tim»l&3-2
Test Data • Vehicle No. CR-08
(Test — I
k*e I— M240 — I I— ASM —I
HC CO HC CO HC CO NO* HC CO NO*
MB * met ft IM IM «M *M »M aM
MMflt '•r'Mre 2771 M L£ S7 O3I JJL 124 2J1 0,4* S.7
1AW9J SfettTe* — 40 044 H« 0.7» _ _ _ _ _ _
1/I4/M LM 0*240 2977 MOCIlUaOTLil 4.1 2&OL333J 1.63
1/26W LM 04240 3161 7S 03t 41 O.M Ltt 4.4 101 IM240 Ms actually increased slightly from the first test to the
last.
The) moat peaaimlatlo scanario on thia vehicle ia that one* the vehicle
failed the Arizona teat for CO* the) mechanic maladjusted the vehicle, took it
to an X/M Ian** where it passed, and then undid the maladjustments. These
undid maladjustments were then observed on the January 14 Arizona retest. A
more benign conclusion ia that the mechanic performed incomplete repairs, but
84
-------
the repairs were ultimately sufficient to pass the Arizona test. Also, th«
repairs were sufficient to pass the ASM NOx outpoints (HC and CO were always
below the ASM outpoints), while they were not sufficient to pass the ZM240 NOx
outpoint, nor were they sufficient to pass the ZM240 HC cutpoint. In fact,
the ASM would not even have identified this vehicle as a high HC emitter.
Clearly, on this particular vehicle, conraercial repairs were not
sufficient to pass the IM240, but were sufficient to pass the ASM test.
Reviewing the data for all 17 vehicles that failed the initial ASM test,
all of these vehicles, except CR-07, CR-21, and CR-22 eventually passed the
Arizona HC I/M retest. Also, all vehicles, except these three vehicles and
CR-06 eventually passed the Arizona CO I/M retest.
However, vehicles CR-03, CR-04, CR-06, CR-08, CR-09, CR-10, CR-15, and CR-
16 which initially failed the IM240, continued to fail HC on the IM240 after
all conmercial repairs (see figure 5.6.3-1). Further, after all connercial
repairs, these same vehicles passed the ASM HC cutpoint (note CR-8, CR-9, and
CR-15 passed the initial ASM test, see Figure 5.6.3-1). Given the
similarities of the ASM and the Arizona test, these data suggest that the
level of HC repair on the ASM would be similar to the current Arizona I/M
test. This assumption oa test similarity and stringency of repair
effectiveness is further supported by the fact that the three vehicles that
failed to pass the Arizona test after repairs (CR-7, CR-21, and CR-22) were
also the only vehicles that failed the after-repair ASM test (see Table Figure
5.6.3-1).
Another method of looking at the ability of the ASM to enforce proper and
sufficient repairs is to look at the test status of the ASM results before and
after repairs relative) to the before and after IM240 status. The test status
for the 17 vehicles) initially failing the ASM for at least one pollutant is
listed in a truth-table) forsmt in Table 5.6.3-3 by pollutant (i.e., HC, CO,
and NOx). The) roughly sqoara boxes in a diagonal row represent tost results
where tha ZM240 and MM status before and after repair were identical.
Deviations) frozfrth* diagonal row, obviously represent results where the status
differs betveeetths) XM240 and ASM. The fuzzy horizontal rectangular box in
Table 5.6.3-*highlights those vehicles which passed the ASM test after
repair, but vare* still failing the ZM240 for HC, CO, or NOx.
Frost the Table, a total of 11 vehicles continued to fail thsj IM240 HC
cutpoint aftar repair. Of these 11 vehicles, 8 vehicles (or 73%) passed the
ASM aftar repair (threa of the eight also passed the initial ASM tast). All
eight vehicles also passed the Arizona HC cutpoint after repair. As
previously mentioned, in the) 11 vehicle sample that continued to fail the
IM240 HC cutpoint, 100% of the vehicles that failed the Arizona HC cutpoint
85
-------
(CR-07, CR-21, and CR-22), also failed the ASM HC cutpoint. Thus, every
vehicle that continued to fail the Arizona HC cutpoint also continued to fail
the ASM HC cutpoint. In other words, in this sample of commercial repairs,
the ASM did not fail anymore retest vehicles than the Arizona Z/M test.
CO test status represents somewhat of a mixed bag. Seven vehicles
continued to fail the 1M240 for CO after repair. Only one vehicle in this
group of seven (or 14%) passed the ASM for CO after repair. This vehicle also
passed the Arizona retest. In this seven vehicle sample that continued to
fail the ZM240 CO cutpoint, six vehicles continued to fail the ASM, and four
vehicles (CR-06, CR-07, CR-21, and CR-22) failed the Arizona retest. In this
case, the ASM found 2 more vehicles than the Arizona I/M retest after
commercial repairs.
However, it should be noted, that four other vehicles had anomalous CO
results. Two vehicles, CR-15 and CR-16 failed the initial IM240 for CO,
passed the initial ASM, and subsequently passed both the IM240 and ASM for CO.
Two other vehicles (CR-09 and CR-25), passed the initial IM240, failed the ASM
for CO, and also subsequently passsd both the IM240 and ASM reteats. If the
ASM was as good as the XM240 in identifying vehicles that should fail a retest
(at the same overall failure rate), one might expect a random scatter on each
side of the diagonal boxes, particularly for vehicles just marginally failing
or passing (which all of these were, except CR-25). Even so, all of these
vehicles also passsd the Arizona CO retest. So they do not represent any
additional retest failures following commercial repairs that the ASM would
have found over the standard Arizona test. Also note, that these four
vehicles initially failed and continued to fail the IM240 for HC or NOx, and
-that the commercial repairs reduced the IM240 CO levels in all cases.
A total of 5 vehicles ia this sample of 17 continued to fail IM240 NOx
after commercial repairs. Two of the five (or 40%) passed the ASM after
repairs. The) Arizona Z/M tost doss not teat for NOx, therefore, it is more
difficult to judge) thm effectiveness of the ASM (using the Arizona test as a
surrogate) to fores) proper and sufficient commercial repairs.
This analymis) began with a concern about the ability of the ASM test to
foster propss> and sufficient commercial repairs following aa Z/M failure.
Because of thm general similarity of tho Arizona Z/M tost to thm ASM, it was
expected that repairs) targeted by the) commercial repair industry towards the
Arizona test would bo similar to those that would be targeted towards the ASM,
at least for HC and CO. Thus, if the ASM were more effective ia forcing
better repairs than ths> Arizona tost, tho ASM retest should fail more cars for
a given pollutant than thm Arizona retest. Further, if tho ASM were very
effective in forcing proper repairs, it would fail as many ears, for a given
pollutant, as aa ZM240 rstsst.
36
-------
The analysis shows that for this small sample, the ASM fails no more cars
for HC after commercial repairs than the Arizona retest, and fails only
twenty-seven percent of those that failed HC on the IM240 after repair. These
results imply that the ASM test would not force the repair industry to make
any more.repairs for high HC emissions than the current Arizona test, and
obviously, not as many HC repairs as the ZM240. Therefore, the data from this
sample suggest that the repair effectiveness credits for HC in the MOBILE
model for commercial repairs on the ASM should be no greater than that
currently given for existing basic I/M programs.
The analysis for CO retest failures, indicates that the ASM found two more
vehicles than the Arizona test. The Arizona I/M retest found about 57 percent
of the IM240 retest failures, and the ASM found about 86 percent of the retest:
failures. Thus in this small sample, it appears that an ASM retest would have
forced the repair industry to make additional CO repairs over and above those
that would have been required to pass the Arizona outpoints, but again, not as
many as the IM240 outpoints would require. These results suggest that the
repair effectiveness credits for CO in the MOBIL* model for commercial repairs
on the ASM should probably be given additional credit over that currently
given for existing I/M program*. The additional credit would be approximately
equal to €0 percent of the) difference between that currently given for
existing I/M programs and that given to I/M programs employing the IM240.
However, given the potential ease) that unscrupulous mechanics could defeat the
CO portion of the ASM retest, assigning additional CO repair effectiveness
credits in the model for ASM over those currently given for existing I/M
programs would be difficult to rationalize at this time.
The analysis fox MOx retest failures is somewhat hampered by the fact that
the Arizona test only fail* vehicles for HC and CO. Iven though the repair
industry was not repairing vehicles; to an MOx standard, the ZM240 and ASM
retests after commercial repairs can be used to determine whether either
retest would hav« forced the repair industry to make additional repairs.
Clearly, beta testa would have required some vehicles to get additional
repairs fo» high NOx emisaiona. However, the results from this sample
indicate thafc as> ASM retest would only require (0 percent of the vehicles that
failed the O04t retest to get additional NOx repairs. Therefore, this result
would suggest that the repair effectiveness credits for MOx in the MOBIL*
model for commercial repairs on the ASM should be about only sixty percent of
that given for the IM240.
87
-------
IM24O
HC
CO
NOx
HC
CO
NOx
HC
CO
NOX
HC
CO
N0»
Table 5^3-3
Effect of Commercial Repairs on Test Status
ASM Status
(Statu* B«£ox« Repair - Statu* Aft«r Rapalr)
CR-07.CR-21.
OI-22
CR-04.CR-OZ
Ot-21.0t-22
CR-9.OM3,
Ot-25
CR-03.CR-04,
CR-06.CR-10,
CR-16
CR-10
aux.ai-1*
Ot-05,OU13
CR-01.CR-05,
CR-13.CR-26
Ot-02
au».at-2S
OI-08.CR-09,
CR-15
CR-01.CR-26
CR-15. CR-16
ai-
-------
S.C.4 One-Mod* Repairs on ASM
The objective of thia analysis was to investigate the theoretical effects
of targeting the ASM repairs to a single mode. That is, if it were possible
for a mechanic to reduce the emissions sufficiently on a single mode while
leaving the remaining three modes unaffected:
• Could an ASM failing vehicle, with such a repair, be made to pass the
ASM composite outpoint?
• What are the emission characteristics of such passing vehicles?
Examining the 106 laboratory test vehicles, we can determine whether the
as-received NO* emissions met or exceeded an FTP NOx standard of 2.0 g/mi.
Also, we can determine FTP HC/CO emission range. That is:
• Pasa
FTP HC $ 0.41 and CO * 3.40 (g/mi)
• Marginal (Failing) Emitters
FTP HC > 0.41 or CO > 3.40 (g/mi)
and
FTP HC i 0.82 and CO i 10.20 (g/mi)
• High Emitters
FTP HC > 0.82 or CO > 10.20 (g/mi)
and
FTP HC S 1.C4 and CO S 13.60 (g/mi)
• Very High Emitters
FTP HC > 1.64 or CO > 13.60 (g/mi)
and
FT* EC S 10.09 and CO £ 150.00 (g/mi)
Bitter*
: > 10.00 or CO > 150.00 (g/mi)
Claaaifying the) laboratory vehiolea in thia way producea tea strata/
however, two of thoM strata arc «qpty, and one stratum has only • single test
vehicle. Uaing the weighting factors (Table S.2.5.2), we) eaa model the lane
vehicles and characterise the) emissions of that simulated lane sample of 2,071
1983 and new fuel-injected passenger ears. (Actually, the lane sample was
2,070 ears* the, additional vehicle resulted from rounding off the) estimated
89
-------
number of vehicles in those eight strata.)
following table.
The distribution is given in the
NOXS2J
NOx>2J>
Laboratory Sample
FTPHC/COEmlsetone
Pees
27
0
Meroinel
36
4
HUM
10
3
V.Htah
18
7
Super
1
0
Simulated Lane Fleet
Piea
808
0
FTP HC/CO Emleelene
Uiralnel
934
24
Won
98
18
V.Htah
143
42
Super
6
0
An ASM outpoint of 1.00/8.0/2.0 (i.e., composite ASM HC £ 1.00, composite
ASM CO S 8.Or and composite ASM NO* S 2.0) will fail 372 vehicles in that
simulated lane fleet. The distribution of those 372 vehicles is given in the
following table.
NOXilO
NO»»2J
QO/1
FTP HOCO Emleetone
12
Super
54
FTPHCCOfmlMlofw
V
•
144
18
30
12
72
36
Super
6
If mechanics were able to repair those 372 vehicle* so that the emissions
on the 2525 mode, the 50 mph mode, and the idle mode remained unchanged, but
.(he emissions (HC, CO, and HOx) oa the 5015 mode were reduced by 80 or 90
percent (the model yields the same result for each), then only 42 of those 372
would be able to pass the> 1.00/8.0/2.0 outpoint. Thus, a repair strategy that
targeted only the) 501S mod* would result in "successfully* repairing only
about 11 percent of the) originally failing vehicles. The distribution of
those 42 passing vehicles is given in the following table.
90
-------
Passing after Reducing 5015 Mod* by 10 or 90%
NOXS24
0
0
Laboratory 81
FTPHG
R/mfBinsI
1
0
vomit
HWw
0
0
unpls
Mien*
V« Hlflll
0
0
Super
0
0
PtS*
0
0
Simula!
FTPHC
MWOjnsI
42
0
•dLHM
COEml
Htah
0
0
> Fl«*t
Mton*
V.Htoh
0
0
Super
0
0
Rather than attempting to reduce the emissions on the 5015 mode by a flat
percentage* the mechanic could target the typical emissions on the 5015 mode
of vehicles whose ASM composite emissions pass the 1.00/8.0/2.0 outpoint.
Such a repair strategy would not change a single failing vehicle into a
passing vehicle in our model.
If mechanics were able to repair those 372 vehicles so that the emissions
(HC, CO, and HOx) on th* 2S2S mod* were reduced by 80 percent (while th*
emissions on the other thre* modes remained unchanged), then only 30 of thos*
372 would be able to pas* th* 1.00/8.0/2.0 outpoint. Th* distribution of
those 30 "successfully* repaired vehicles is given in th* following table.
•ta* i%4«4»^4tmfe4.«
Passing after Reducing 252S Mod* by 80%
NOKȣ0
V.
Super
12
Super
Reducing th* 2525 mod* emi**ion* by 90 percent would add 41 vehicles (42
marginal HC/CO emitter* with NQx 1 2.0 and 6 very high HC/CO emitters with
NOx > 2.0) to th* 30 who** estimated ASM composite score would pass th*
outpoint. 7b**» • repair strategy that targeted only th* 2525 mod* would be
successful, omUaly 21 p*xc*at of th* originally failing vehicle*.
at*
Rather them attempting to reduce the emission* oa the) 2S2S mod* by a flat
percentage, th* mechanic could target th* typical emission* oa th* 2525 mode
of vehicle* who** ASM composit* emissions pass th* 1.00/8.0/2.0 outpoint.
Such a repair strategy would not Chang* a single failing vehicl* into a
passing vehicl* in our mod*!.
If mechanics wer* able to repair those 372 vehicle* so that th* emissions
(HC, CO, sad HO*) oa th* 50 mph cruise mod* w«r* reduced by 80 percent (while
91
-------
the emissions on the other three modes remained unchanged), then 264 of those
372 would be able to pass the 1.00/8.0/2.0 outpoint. The distribution of
those 264 passing vehicles is given in the following table.
Fla^fe niafcfibutlon Tailing 1 QQ/i Q/2 Q
Passing after Reducing 50 mph Cruise Mode by 80%
NOx>24
Laboratory Sample
FTP HC/CO Emissions
Pas*
3
0
MeVQIfWl
5
2
IMialisi
mynv
3
2
V.Htah
3
2
Super
0
0
Simulated Lane Fleet
FTP HOCO Emissions
Pass
54
0
• f^-,,1 |
•"•WgsTMef
138
12
High
18
12
V.HIgh
18
12
Super
0
0
Reducing the 50 mph cruise emissions by 90 percent would add 6 vehicles (all
with very high HC/CO emitters and NOx > 2.0) to the 264 whose estimated ASM
composite score would pass the cutpoint. Thus, a repair strategy that
targeted only the 50 mph cruise mode would result in "successfully* repairing
about 73 percent of the) originally failing vehicles.
Rather than attempting to reduce the emissions on the 50 mph cruise mode
by a flat percentage, the) mechanic could target the) typical emissions on the
50 mph cruise mode of vehicles whose ASM composite emissions pass the
1.00/8.0/2.0 cutpoint. Such a repair strategy would result in "successfully"
repairing 162 (44%) of the) originally failing vehicles. The distribution of
those 162 passing vehicles is given in the following table.
yia^fe Pt«»flhn»ia« railing 1 QQ/t 0/2 Q
Passing after ftedaoing 50 mph Cruise) Mod* to Meaiaal Seer*
NOx i 10
NOx»laV
If IMnh
T. nign
Super
42
rTF HC/CO Inhesions
V.
84
24
12
Super
It menlmfntr v*re> able) to repair those 372 vehicle* so that the) emissions
(only HC and 00} OB the> idle) mode were reduced by 80 or 90 percent (the model
yields the) ssaa> result foe each) while the emissions oa the other three modes
remained unchanged/ them only 9C of those 372 would be able to pass the
1.00/8.0/2.0 cutpoint. Thu*> a repair strategy that targeted only the idle
mode) would result in "successfully* repairing only about one-fourth of the
originally failing vehicles. The distribution of those) 9C passing vehicles ia
given ia the) following table).
92
-------
1-QQ/-Q/2.O
Passing after Reducing Idle Mod* by 80 or 90%
NOx>24
Pass
0
0
Laboratory Si
F1PHG
levokwl
2
0
ICOBflt
UbrfM
niflnv
1
0
rnpls
MM!)*
V. Htah
1
0
Super
0
0
Pis*
0
0
Slmulat
FTPHC
mAW^MfemAft
MVQV1M
84
0
•dLam
CO fin
UbeJt
nion
6
0
»PlMt
MlOTM
V.HIflh
6
0
Super
0
0
Rather than attempting to reduce the emissions on the idle mode by a flat
percentage, the mechanic could target the typical emissions on the idle mode
of vehicles whose ASM composite emissions pass the 1.00/8.0/2.0 outpoint.
Such a repair strategy would result in "successfully* repairing only 60 (16%)
of the originally failing vehicles. The distribution of those 60 passing
vehicles is given in the following table.
a Tatlia 1 QQ/t.Q/2 fl
Passing aftes Kedveing Idle Mode) to Woainal Scor*
NO»24
FTPHOCOI
V. HlflH
Super
FIFHC/COImtaelorw
42
12
Super
Tightening the ASM outpoint from 1.00/8.0/2.0 to a more stringent outpoint
of 0.40/8.0/1.S produce* similar results in our model.
An ASM outpoint of 0.40/1.0/1.5 will fail 587 vehicle* in that simulated
lane fleet. Tha distribution of those 587 vehicle* is given in the following
table.
A* with th* 1.00/8.0/2.0 outpoint, single-end* repair* that reduced the
5015 mod* emission* by 80 or 90 percent would succeed in "successfully"
93
-------
repairing only 108 (18 percent of the 587 failing vehicles)
of those 108 vehicles is given in the following table.
The distribution
ni •«•«•**»««; ion railing 0.40/8 ft/1 S
Passing after Reducing 3013 Mode by 80 or 90%
NOxiiO
NO»2J
1
0
Labori
FTPHC
•ML*eiMa»feAl
•NsWglTW
5
0
rtory Si
CO Em
***-*--
rwyiM
0
0
unple
V. Woo
0
0
Super
0
0
Pass
6
0
Simula!
FTPHC
Marginal
102
0
ad Lane Rest
"CO Emi
Htah
0
0
•won*
V. High
0
0
Super
0
0
Rather than attempting to reduce the emissions on the 3013 node by a flat
percentage, the mechanic could target the typical emissions on the 5013 mode
of vehicles whose ASM composite emissions pass the 0.40/8.0/1.5 outpoint.
Such a repair strategy would result in "successfully* repairing only 54 (16%)
of the originally failing vehicles. The distribution of those 54 passing
vehicles is given in the following table.
Q 1Q/S
Passing aftec Reducing 3015 Mod* to Vomlnal Seer*
NOXS24
NOX»24
FTP HC/CO Emissions
V. Hlflli
Super
FTP HCVCO Emissions
V
. T«
48
If mechanics were able to repair those 387 vehicles so that the emissions
on the 2525 mod* were reduced by 80 percent (while) the emissions on the other
three mode* remained unchanged), then only 138 of those 587 would be able to
pass the 0.40/8.0/1.3 outpoint. The distribution of those 138 vehicles is
given in the) following table).
94
-------
y^pjft* ptaferttmfctan >allln«y Q ^Q/g ft/T jt
Passing aftax Reducing 2525 Mode by 10%
NOx»24
FTP HC/CO Emissions
V.Mgh
9
1
Super
Slmulstsd Lane Rest
FTP HC/CO Emissions
Mugjral
126
V.HIgh
6
Sui
0
Reducing the 2525 mods; emissions by 90 percent would add 12 vehicles (6
passing HC/CO with NOa 1 2.0 and € marginal HC/CO emitters with NOx > 2.0) to
the 138 whose estimated ASM composite score would pass the outpoint. Thus, a
repair strategy that targeted only the 2525 mode would be successful on only
about 26 percent of the originally failing vehicles.
Rather than attempting to reduce the emissions on the 2525 mode by a flat
percentage, the mechanic could target the typical emissions on the 2525 mode
of vehicles whose ASM composite emissions pass the 0.40/8.0/1.5 outpoint.
Such a repair strategy would result in "successfully* repairing only 102 (17%)
of the originally failing vehicles. The distribution of those 102 passing
vehicles ia given in the following table.
0 .4Q/t .0/l.S
Passing aftex ftedueing 2525 Mods) to Vominal Score,
NOx s 2.0
FTP HC/CO I
V.HiQh
Super
96
V.Htah
Super
If mechanic* wer« eblsj to repair those 587 vehicles so that the emissions
on the 50 mpb exvia* mod* were reduced by 80 percent (while the emissions on
the other thne> sadee remained unchanged), then 3€5 of those 587 would be able
to pass the>».40/§.0/1.5 outpoint. The distribution of those 365 passing
vehicle* ia> gins* ia tte following table.
95
-------
fit
Failing 0.10/8 0/1 S
Passing aft** Reducing SO aph Cruise Nod* by 80%
NOXS2J
Laboratory Sample
PMO
6
0
FTP HOCO Eml
Marginal
12
1
HUM
4
0
aalone
V* Hlflll
2
0
Super
0
0
Simulated Lano Float
Paao
72
0
FTP HOCO Emteelons
Marginal
251
6
IHnli
man
24
0
V. Htah
12
0
Super
0
0
Reducing the 50 nph cruise emissions by 90 percent would add 12 vehicles (all
with NOx > 2.0; 6 of which with marginal HC/CO and 6 with high HC/CO) to the
365 whoso estimated ASM composite score would pass the outpoint. Thus, a
repair strategy that targeted only tho 50 nph cruise mode would result in
"successfully* repairing about €4 percent of the originally failing vehicles.
Rather than attempting to reduce the emissions on the 50 mph cruise mode
by a flat percentage, th* mechanic could target tho typical emissions on tho
50 mph cruise mode of vehicles whoso ASM composite emissions peso tho
0.40/8.0/1.5 cutpoint. Such a repair strategy would result in "successfully'
repairing only 275 (47%) of tho originally failing vehicles. Tho distribution
of thoso 275 passing vehicles ia given in tho following table.
0.4Q/a fl/l.S
Passing aftox Reducing 50 mph Cruise) Mod* to nominal Score
NOX »10
V.HIoh
Super
48
Simulated Lane Pleat
FTP HOCO Emission*
V Utah
w» I Iflill
191
24
12
If mechanic* w*r* abl* to repair thoso 587 vehicle* so that tho omissions
on th* idl* mod*) w*r* roduced by 80 or 90 percent (tho model yiolda tho same
result for each} vail*) th* omission* on tho other thro* modes remained
unchanged, thoatoaly 42 of thoso 587 would bo able to paaa th* 0.40/8.0/1.5
cutpoint. Ttaav • ropaix strategy that targeted only th* idl* mod* would
result ia •succ*aafully* repairing only about seven percent of th* originally
failing vohicl**. Th*) distribution of those 42 passing vehicle* is given in
tho following tabl*.
9<
-------
Fallin Q.AO/i.Q/1 Jt
tassiag after Reducing Idle Mod* by SO ox 90%
Pie*
0
0
Laboratory Si
FTP HO
Itaratna!
1
0
rcocmi
Htahs
0
0
mpto
W Utah
»• nvn
0
0
Super
0
0
PaSO
0
0
Sbmilat
FTPHC
" • •
NeWuWtv
42
0
•dLam
COImJ
Htah
0
0
I Fleet
•stone
V.HIgh
0
0
Super
0
0
Rather than attempting to r«duc« th« •missiona (only HC and CO) on the
idle mod* by a flat percentage, the mechanic could target the typical
emissions on the idle mode of vehicles whose ASM composite emissions pass the
0.40/8.0/1.5 outpoint. Such a repair strategy would have produced exactly the
same result (i.e., 42 passing vehicles) as would reducing the idle emissions
by a flat 80 or 90 percent.
From the preceding two examples (i.e., using outpoints of 1.00/8.0/2.0 and
0.40/8.0/1.5), the only potentially effective "single-mod* ASM repairs" are
those repairs targeted at the 50 mpb cruise mode (reducing emissions by 90%).
However, the model predicts that those repairs would not be successful on 27
to 36 percent of the originally failing vehicles. The distributions, of those
vehicles that ar* still failing th* respective ASM outpoint after repairs
targeted on th* 50 mph cruis* mod* (reducing emissions by 90%), are given in
th* following table.
0.40/1 o/i.a
Cruise Mod* by 90%
After Reducing 50
NOX»2*
SIU Ming Outpoint of 0^40/10/1 J
Super
6
Super
6
„ w* can see that th* vehicle* that th* model
predict* will continue) to exceed th* respective ASM outpoints, even after
single-mod* repairs targeted at th* 50 mph cruis* mod*, ar* among th* highest
emitters in th* simulated Ian* fleet.
97
-------
5 . 7 Purge Analyses
S. 7.1 Introduction
In purge testing, the concern is not that too many malfunctioning vehicles
will pass a teat. Instead/ the major concern ia too many properly functioning
vehicles will fail a test that attempts to replace a test with real-world
driving behavior with a few steady-state modes. These steady-state modes may
not provide some vehicles with the opportunity to purge. Thus, the purpose of
this section was to compare the canister purge system false failure rates
(errors-of-commission or Bcs) for the ASM and IM240. The data indicate that
the ZM240 is significantly less likely to falsely fail vehicles for purge than
the ASM.
Vehicle evaporative emissions contribute significantly to the VOC
inventory. Because vehicle fuel tanks and carburetors must vent to atmosphere
for proper vehicle operation, cartoon canisters are added to collect
hydrocarbon molecules which would otherwise escape. Because the carbon
canister has a finite capacity, which if exceeded, allows hydrocarbons to
escape, the canister must be kept purged of stored hydrocarbon molecules. The
evaporative control system includes a purge system which draws stored
hydrocarbons into the engine where they are burned.
Most properly functioning canister purge system* do not purge constantly;
instead, most only purge; when their CCM computer algorithms call for purge.
Driveability or emission problem* accompany purge that initiates during
unfavorable conditions, so purge algorithm* are designed to take advantage of
opportune condition*. The purge algorithm* are known to vary widely from
model to model. So, the main problem for an I/M teat is to provide vehicle
operation that will coincide with the conditions necessary to induce the
system to activate) canister purge. EPA ha* found that some vehicle* only
purge during acceleration* or deceleration*, which is problematic for ateady
state testa such •• the) Aflat and could result in falsely failing vehicles with
purge system* that are) properly functioning.
Also, aeaafevehielea have) timer* that don't allow purge for several minutes
after the) eofl*** i* ataxtad or * specified operating temperature ia reached,
so all el*e> being equal* tha longer the teat duration, the lower the
probability of purge) falaa failure*. This also make* tha teat order
important, sinca tha taat that vaa performed second i* more likely to achieve
purge than tha initial taat. Thia ia one reason tha teat procedure in our
study ia Maaa required tha taat order to be reversed each time another car was
tested and why tha angina waa raatartad just before each taat.
98
-------
The ZM240 purge flow was summed over the full ZM240. The ASM purge flow
was measured and sunned over the full four nodes of the ASM including
transient segments of the ASM cycle, in contrast to the ASM exhaust emissions,
which were only measured during the four ASM steady-state modes. Because the
flow measuring equipment is the same for both transient and steady-state
tests, it was practical to measure purge flow on the ASM during the three
accelerations and the single deceleration needed to complete the four ASM
modes.
5.7.2 The Database
The database for the lane purge analysis was restricted to vehicles that
met all of the following criteria:
- as-received purge data were available for both the XM240 and the ASM,
- the test order was known,
- and the data passed the purge QC criteria (see Appendix C).
The resulting database consisted of 1170 vehicle*. Of these, 577 received
the ZM240 first and 593 received the ASM test firat. The comparisons made for
this analysis included failure rate comparisons, and comparison* of vehicles
for which the ASM and XM240 purge) status did not agree, or "false failures".
in addition, these comparisons were made for data stratified by test order.
The standard used for these comparisons was 1.0 liter/test.
5.7.3 Th* Reaulte
The results of the failure rate comparisons are as follows:
• Overall purge) failure rate*:
- The XM240 failed 7.43% (87 vehicles).
- The ASM failed 11.45% (134 vehicles).
• Initial test failure ratea:
• Tbei IM240.1st failed €.93% (40 vehicles).
AM. 1st failed 11.13% (6C vehicles).
• Second teat failure rateas
• The) DC40.2nd failed 7.92% (47 vehicles).
- The ASM.2nd failed 11.79% (68 vehicles).
These higher failure ratea for the ASM raise the question of whether the
ASM correctly identified non-purging vehicles that the ZM240 missed, or
whether the ASM incorrectly failed vehicles. Passing the DC40 purge test
requires either that purge actually occurs or that the) measurement system
99
-------
falsely indicate that purge is occurring. Sine* the ASM and IM240 were run
with the same measurement system, and results were reported electronically
without human intervention, it is not conceivable that a measurement error
made some cars pass the ZM240 and fail the ASM. Consequently, the ASM-
fail/!M240-pass cars must be considered improper fails by the ASM, and vice
versa, with a possibility that test order was a contributing factor in
specific cases despite the engine restart for both tests. However, since the
sample has essentially an equal number of each test order, test order should
not be a relevant factor overall.
The next set of statistics implies that both the ASM and the IM240 falsely
fail vehicles, but the ASM falsely fails more vehicles.
• Overall false failure rates (fails one test but not the other):
- 1.1% or 13 vehicles failed the ZM240 but passed the ASM.
- 5.13% or 60 vehicles failed the ASM but passed the IM240.
• False failures on initial test:
- The IM240.1st falsely failed 1.21% (7 vehicles).
- The ASM.1st falsely failed 4.22% (23 vehicles).
• False failure* on second test:
- The ZM240.2nd falsely failed 1.01% (6 vehicles).
- The ASM.2nd falsely failed €.07% (35 vehicles).
Figure 5.7.1 graphically illustrates the comparison of false failure
rates.
100
-------
7.0% • •
6.0% ••
5.0% • >
4.0% •>
3.0% •-
2.0% • -
1.0% ••
0.0%
Overall
N-1170
5.1%
Figure 5.7.1
Comparison of False Failure Rates for
ZM240 purge vs ASM purge
ASM.2nd
N-577
1.1%
ASM. 1st
N-593
4.2%
ZM240.1st
N-577
1.2%
Overall - False
Failure Rate
Initial - False
Failure Rat*
6.1%
IM240.2nd
N-593
1.0%
9 recondit ioned
False Failure
Rate
DASM Purge - False Failure Rat* fl ZM240 Purge - False Failure Rate
The moat relevant comparison is the initial tests (ASM. 1st • ZM240.1st)
because they are more representative of the conditions and vehicle
preconditioning expected in official I/M programs than the preconditioned
tests. The purge results for the initial tests were similar to the overall
results; the ASM. 1st'a false failure rate was 3 percentage points higher than
for the IM240.1st.
With a 3 to 4% falsa) failure rate* the ASM purge test could cause severe
problems to I/M programs) la the form of frustrated consumers and skeptical
mechanics.
As disease** in the introduction, test order was expected to be important
because th* tee* that we* performed second would be more likely to achieve
purge than the teat that was performed first. Contrary to expectations,
however, the ASM.2nd exhibited a 0.63% increase in failure rate and a 1.85%
increase in false failures when compared to the ASM. 1st. The the IM240.2nd
also produced a higher failure rate (+1.0%) than the DC40.1st, but the
lM240.2nd's false failures decreased 0.2% compared to the XM240.1st.
In addition, the false failure rate for the XM240.2nd is markedly better
than for the ASM.2nd when viewed as a percentage of failures. Figure 5.7.2
101
-------
shows that the false failure rat* dropped from 17.5% of the failing IM240.1st
vehicles to 12.8% of the failing ZM240.2nd vehicles. In contrast, the false
failure rate increased to 51.5% of the ASM.2nd failing vehicles from 37.9% of
the ASM.1st failing vehicles. So although the false failure rate for both
tests is expected to decrease further if the engine restart is avoided before
performing the second-chance test, these data suggest that retesting is more
effective in reducing ZM240 false failures than ASM false failures.
Figure 5.7.2
raise Failure Rates as a Percentage of Failures
for XM240 purge vs ASM purge
Overall - False
Failure Rate
Initial - False
Failure Rate
Preconditioned
- False Failure
Rate
OASM Purge-Fals* Failure Rat* fl IM240 Purge-False Failure Rate
Overall, 74 vehicle* failed both th* ASM and IM240. Th* ASM falsely
failed 60 additional vehicles while the IM240 falsely failed only 13
additional vehicle*. A* shown in Figure 5.7.2, 44.8% of th* 134 vehicles
failing they AaMpurg* wer* false failure* compared to 14.9% of th* 87 vehicles
failing th*rXltttt purg*. In addition, 37.9% (29 of
-------
failures. However, it is speculative whether this will significantly reduce
the ASM false failure rate. Also, second-chance testing adds cost. Since the
false failure rate increases from the ASM.1st to the ASM.2nd and decreases for
the IM24.2nd compared to the ZM240.1st (See Figure 5.7.2), the data indicates
that second-chance testing may not be as effective for the ASM as for the
IM240.
Second-chance testing costs will be lower for the ZM240 because it fails
fewer cars initially, thus, requiring fewer retests and some vehicles just do
not purge during steady-state operation. For these vehicles, an alternate
cycle such as the IM240 would be required. Dynamometer costs would then
increase because inertia simulation is needed, but the more expensive IM240
exhaust measurement systems would not be needed.
In summary, the comparison of ASM purge and XM240 purge shows that the
IM240 is superior in correctly identifying vehicles with malfunctioning purge
systems, with false failure rates of 4 to 6% for the ASM (3 to 6 times higher
than IM240 false failure rates), an additional second-chance test will be
required. And since some vehicles simply do not purge on the ASM steady-state
modes, even with purge) measured during the accelerations between modes, an
alternate cycle such as the ZM240 may be required for retests. In conclusion/
the ASM purge test is substantially less effective than the ZM240 purge test.
5. • ZM240 Improvements and the Four-Mode ZM240
The purpose) of this section is to convey that refinements are possible
which would make the XM240's performance a "moving-target," and to further
reiterate why one sample should bet used to develop the ASM-mode weighting
factors and an independent sample used to evaluate the ASM's effectiveness.
EPA's recommended ZM240 outpoints of 0.80/15.0/2.0 + 0.50/12.0 represent a
compromise between failing high emitting vehicles and not failing clean
vehicles. fts outpoints ax* tightened* the iDRa generally increase at the
expense of inenasing the) possibility of errors-of-commission. Increasing the
power of the) tawe (i.«., the ability to distinguish between malfunctioning and
properly fTimtqiiirg vehicles) serves the public good, in that the high
emitters not identified by the) test are not repaired* so the cost of testing
such vehicles is not rewarded by air quality improvements that accrue from
identifying and repairing such vehicles. More tangible is that vehicle owner
satisfaction and acceptance) of I/M programs increase with lower errors-of-
coonission.
EPA is not content foe all tias with the absolute performance of the ZM240
as now defined. Although ia a relative sense, its performance is superior to
103
-------
. Jfc.
any of the alternative I/M tests, it can be improved. Consider for example the
IM240's 15 g/mi CO outpoint. This is more than four tinea the FTP CO
standard, and unlike the FTP which includes a cold start9, the IM240 is to be
performed on fully warmed up vehicles. So the errors of omission (vehicles
which pass, but should not) are higher than if a more stringent CO standard
were used. EPA testing has shown that tighter outpoints will identify more
high emitters, but also fail some properly functioning vehicles. Although the
IM240 is considerably better than any alternative I/M test, in this regard,
there is no question that its performance can be improved. The IM240's
performance can be improved in two areas:
- Reduce the teat-to-test variability so that outpoints can be tightened
without falsely failing clean vehicles.
- Us* statistical techniques to improve the IM240rs correlation with the FTP.
Such improvements will serve the public interest by increasing the air
quality yield per teat-dollar, so alternative testa should be evaluated
againat the state-of-the-art of IM240 testing rather than the IM240
performance, aa it existed, when the I/M Rule was published. Proponents of
alternative I/M teata nay point out that if the XM240'a performance, aa it
stood in November 1992, waa good enough to meet the performance standard, then
this performance standard should be the standard for alternative teata. While
such a policy may indeed "level-the-playing-field* for alternative testa and
is in fact what la allowed by EPA'a I/M Rule, it ia difficult to argue that
this approach promotes the general welfare and should guide state and local
decision-makers concerned aa much about clean air aa about meeting minimum
requirementa.
S.t.l ftedvoe) Teet-to-Test Variability
Teat-to-teat variability ia the primary reason why the DC240'a outpoints
are so much lea* stringent than the FTPs. The FTP controla a number of
variables that are) widely known to affect a given vehicle'a emiaaiona. Some
variables the*, are) tightly controlled for FTPs were either more loosely
controlled oaxnot controlled in EPA'a ZM240 lane teata. Theae include, among
othera:
9 CO (and HC) eaiaaione are considerably higher during waraup than during
fully warned-up operation.
104
-------
- ambient temperature in the test cell
- humidity in the test cell
- engine temperature (FTP indirectly controls with preconditioning and
ambient conditions)
- catalyst temperature (FTP indirectly controls with preconditioning
and ambient conditions)
- vehicle operation prior to the emissions test (can affect emission
control system timers for purge, air switching, etc., and other
variables affecting emissions)
- evaporative canister loading (FTP indirectly controls with ambient
conditions and vehicle operation during the 12 hours preceding the
FTP emissions test)
- tire pressure
• speed excursions from the nominal speed (±2 mph on FTP vs. driver
discretion for EPA's pilot XM240 testing to date)
- exhaust system backpressure (NOx can be adversely affected by a
constant volume sampler if quality control is not adequate)
- fuel composition
EPA has already made improvements that I/M programs will be required to
implement, but were not implemented during EPA's testing. For example, FTPs
are voided if speed excursions from the nominal speed exceed ±2 mph. In
contrast, much of EPA's data are from vehicles with speed excursions that
exceed ±2 mph. In a committee that included I/M contractors, state X/M
program, officials, XM240 equipment manufacturers, and automobile
manufacturers, a consensus was reached on requiring this tighter speed
tolerance along with additional tighter controls that will reduce test-to-test
variability10.
There are also variables that can not be controlled, such as ambient
temperature and canister loading, but can be compensated for to better
distinguish between malfunctioning and properly functioning vehicles. Given
enough data, computer algorithms can be developed that consider the more
important variables) and apply adjustment factors to the official IM240 test
results.
SiapllstiOY approaches] such as setting tire pressure or providing second-
chance testa fox vehicle* that are within l.S times the outpoints seem costly
to implement because) additional X/M lanes and personnel are needed, but are
judged to be cost effective) since vehicles that should not fail but do, must
be retested after •repairs' anyway.
10 Draft High-Tech Test Procedures, Quality Control Requirements, and
Equipment Specifications, April S, 1993.
105
-------
More sophisticated algorithms utilizing sensors to measure variables like
ambient temperature and catalyst temperature allow relationships to be
developed and used to compute scores. These are more efficient because they
can increase the power of the test without requiring additional I/M lanes and
personnel. Developing these techniques will require substantial data. When
the IM240 is implemented, much data will become available to allow development
of such algorithms.
5.9.2 Statistical Techniques to Improve the) XM240's
Correlation With the FT»
Presently/ the ZM240 score constitutes the sun of the mass emissions
divided by the distance accumulated. Because almost every second of operation
is taken from various segments of the FTP, the two tests correlate better than
any existing alternative I/M test. But their correlation can be improved
using multiple regression. For example, the uncontrolled variables that
attend XM240 testa probably make it appropriate to de-emphasise the initial
operation of the IM240 in computing the score and emphasizing the later
operation. The later operation is somewhat preconditioned by the initial
operation. Also, the IM240 has a higher average speed than the FTP, so de-
emphasising the high speed portion* should produce a better correlation with
the FTP.
The data itself can be used to determine the more appropriate weighting
through the use of regression techniques. For example, EPA divided the IM240
.into four modes as followss
node 1: 0-60 second*
Mod* 2: 61-119 second*
Mod* 3: 120-174 second*
Mode 4: 175-239
A* for thevASM* coefficients are developed by performing a multiple
regression Mtisjuin the) result* from four mode* are the independent variables
and the) m> reevlte 1* the) dependent variable, which allow* the data to
determine the> appropriate; mod* weighting.
EPA tried thi* uciAg the only substantial database with the information
needed (FTP* with ZM240 4 ••ode results or second-by-second result*), which
happen* to be the vehicles on which this report focuses. So the coefficients
had to be developed on the asm set of data to which they were applied. EPA
condemns this practice, as discussed in Section 3.5, but having no
106
-------
alternative, the results are presented only to provide an indication of how
the IM240's performance is enhanced through this approach.
Unfortunately, the legitimate performance increase gained by using
multiple regression to determine appropriate mode weighting can not be
isolated from the inappropriate application of these coefficients to the same
vehicles from which they were developed. So the performance presented in Table
5.8.2 gives an overly optimistic view, but also reiterates that ASM-advocates
who do not accept EPA'a judgement that it is inappropriate to apply
coefficients to the vehicles from which they were developed, should then
compare the ASM performance with inter-linked coefficients to the IM240 also
utilizing inter-linked coefficients.
The multiple regression was performed on the first 91 vehicles in the
database only (this analysis was not repeated when additional data became
available). The results are presented in Tables 5.8.1 and 5.8.2. The
negative coefficients in Table 5.8.1 indicate that insufficient data is
available for developing logical coefficients, which will compensate, to some
degree for the inter-linked performance listed in Table 5.8.2.
Table) 5.8.1
Coefficients Developed from
Multiple Regression of 4-Mode ZM240 vs. FT»
Mode
Constant
1
2
3
4
R*
me
0.03
-0.30
1.16
0.2C
0.09
90.3%
CO
-0.28
-0.18
1.70
-0.01
-0.08
86.0%
M0«
-0.02
-0.07
0.37
0.02
0.5C
69.6%
Tabl« S.tvX illustrate* how the 4-Mode test improve* the tradeoff between
iDRa and Be ntoa at «quivalent failure rates.
107
-------
• •».
Table 5.8.2
4-Modsj ZM240 Performance Veraua Normal ZM240
rail
Rate
12%
13%
14%
15%
19%
20%
23%
I
4 -Mod*
88%
90%
91%
92%
91%
94%
95%
C
Regular
88%
90%
91%
91%
93%
93%
93%
IDRa
CO
4-Mode
65%
66%
66%
66%
72%
72%
73%
Regular
63%
65%
66%
66%
68%
69%
69%
NO*
4 -Mod*
72%
75%
78%
75%
83%
86%
88%
Regular
75%
76%
78%
78%
82%
82%
83%
Ic»
4 -Mode
0.0%
0.0%
0.4%
0.4%
0.4%
0.4%
0.7%
Regular
0.4%
0.4%
0.7%
0.7%
1.4%
1.8%
3.9%
Notice the performance increase in that the iDRs increase and errors-of-
comaission decrease.
In conclusion, developer* of alternative I/M tests should not consider the
performance of the ZK240 to be fixed, while better than any existing I/M
tests, XM240 improvements are possible and desirable. CPA's mission to
improve air quality and enhance the public welfare necessitates evaluating
alternative tests, not against the performance of the XM240 as it was in
November 1992 when the I/M Rule was published, but instead, against the state-
of-the-art.
108
-------
6. Test Programs by Other Organizations
6.1 Colorado Tost Program
Tho Colorado Department of Health (COH) completed an evaluation11
comparing the FTP and the IM240 to the following eight X/M test modes, in the
order the modes were performed:
- 35 mph road load
- 50 mph road load
- ASM 2545
- ASM 2525
- ASM 5015
- Idle tost
- 2500 rpm
Their conclusions included the following:
"The loaded mode [IM240] tests (both [93] second and 240 second]
identify significantly man of the excessively emitting vehicles and man of the
excess eaimkxttthaa do any of ttw steady-state tests. They also have fewer
etnas of commission sod less sensitivity to differences between FTP sod short
test emission levels. Win w ote consideration, eitter the 95 second or the
240 second version of dw QM240] would be the clear choice for the most
accarate sod effective kkntifkation of excessively emitting vehicles.1*
€.2 California Test Program
EPA has received a preliminary analysis12 from the California Air
Resources Board (CAM) comparing the ASM5015 and the ASM2525 to the IM240.
The GARB analyeie looks) favorably on the ASM modee and concludes that the ASM
tests are as effective ae the ZM240. However, there are significant concerns
with CARS's data. The** concerns) include the following:
- They CUB database la not representative of the newer fleet.
11
Ragazzi, et al.
12 Draft Memorandum from Jeff Long, Manager, Analysis Section to Mark
Carlock, Chief, Motor Vehicle Analysis Section, "Comparison of Excess
Emissions Identified by IM240, ASMS015 and ASM252S Tests,* California Air
Resources Board, not dated, received April 15, 1993.
109
-------
- GARB'a tasting is not representative of actual I/M testing.
- All of CARS'a testa were preconditioned.
- GARB'a ASM equations, when applied to EPA's data, demonstrate poor
performance.
EPA is preparing a separate document that will consider CARS's ASM teat
program in more detail.
110
-------
7. Teat Coata Comparison
Supporters of the ASM have frequently suggested that it would be a more
cost-effective test than the IM240, given that the equipment cost is
significantly lower. As Table 5.12.1 shows, the equipment package for the ASM
aeries, with purge and pressure testing, does have a lower total cost than the
IM240, purge, and pressure equipment package.
Table 7.1
Equipment Costs for the ASM Series and IM240
IM240 ASM
£flflt V.epi < pmont-
Pressure Rig $600 Pressure Rig 9600
Purge Meter $500 Purge Meter $500
VDA $1,000 VDA $1,000
Dynamometer $25,000 Dynamometer $20,000
CVS £ Analyzers $79,000" BAR90 fi NOx Bench $19,000
Total $106,100 $41,100
These figures reflect the most recent cost information that EPA has
received from industry. EPA has published previous estimates of the per
vehicle costs of ASM and ZM240 testing in "I/M Costs, Benefits, and Impacts,"
in November, 1992. EPA found, and independent analyses confirmed, that
equipment costs, when spread over the useful life of the equipment, constitute
a relatively small portion of the per vehicle cost of an Z/M test; labor and
-overhead costs are considerably higher. In analysing the) current average per
vehicle inspection cost in s centralised program of $8.50, EPA estimated that
equipment accounted for 219, labor for 96C, 82C went to defray construction
costs, the state oversight fee averaged $1.25, and the remaining $5.26 went to
cover various overhead cost* (for a full discussion of EPA's cost estimation
assumptions and methodology the reader is referred to Sections 5.2 and 5.3 of
"I/M Costs, Benefits, and Impacts,* contained in Appendix M). Current testing
stations have) aa average) peak capacity of 25 vehicles per hour and enough
stations are> constructed to avoid long lines on peak '*••""'< days. Given the
typical pattern of owners' choices about when to cone for inspections, this
results in aa average* actual throughput of 12.5 vehicles per hour which
13 Letter from. Kenneth W. Thomas, Marketing Manager, Z/M Systems, Horiba
Instruments Incorporated, to Bill Pidgeon, U.S. Environmental Protection
Agency, April 7, 1973 and Quotation from Scott P. Corrunker, Salee Engineer,
Combined fluid Products Company to Dan Sampson, U.S. Environmental Protection
Agency, January 27, 1993. These are attached as Appendices J and X.
Ill
-------
translates into 39,000 vehicles per year per lane, and costs are spread over a
multi-year period, five years in most cases.
Throughput is the most critical variable in estimating costs since it
determines the size of the. inspection station network needed for a given area
and the number of vehicles over which costs for each lane are spread.
inspection lanes usually have more than one position, with different parts of
the inspection performed at each one. Hence, throughput is governed not by
the time required to perform the total test sequence, but by the time required
at the longest position. Whether the test sequence consists of the IM240 with
purge and pressure testing or the ASM with purge and pressure testing, the
longest part of the sequence is the tailpipe emissions test.
The combined IM240 and purge test takes approximately three minutes (using
fast-pass and fast-fail) to perform on the average. Allowing an additional
minute to maneuver the vehicle onto the dynamometer and otherwise prepare the
vehicle for testing the total time at the longest position is estimated to be
four minutes. This translate* into a peak lane capacity of IS vehicles per
hour and an average actual throughput of 7.5 vehicles per hour. The ASM
consists of four modes lasting 40 seconds each with a few seconds in between
to change speed. This works out to approximately three minutes per test.
Allowing, again, aa additional minute to maneuver the vehicle onto the
dynamometer and otherwise prepare it for testing, the total test time is about
four minutes, hence, the throughput rates for the ASM is the same as for the
IM240. Average throughput for both tests is 7.5 vehicles per hour. Assuming
that stations operate €0 hours per week, 52 weeks per year, and costs are
spread over a five year period, then equipment costs are spread over a total
•of 117,000 vehicles.
The optimum lane) configuration for both tests is a three position lane
staffed by thre« inspectors. Consequently, as shown in "I/M Costs, Benefits,
and Impacts,* staff, infrastructure and overhead costs are essentially the
same for both tests. The) only difference is in the cost of equipment. Table
5.12.2 show* th« estimated per vehicle costs for performing the ASM and the
IM240. The> eoats are) derived using the same methodology and assumptions as in
Appendix B. Overhead costs for ZM240 and ASM testa are estimated by factoring
the overhead fo* current centralized programs by the change in throughput.
Equipment, and construction costs are obtained by dividing thos« costs over
the total vehicle traffic la a five year period. Staff coata are obtained by
dividing inspectors1 hourly wagea (36.00) by the average number of vehicles
inspected in a hour. State oversight costs are estimated at 91.75 per vehicle
but could vary depending upon the intensity of the state oversight program;
they would not vary between the two test types.
112
-------
Despite the difference between the coats of the equipment packages
required for the two tests, the total cost per vehicle, factoring in all
necessary costa involved in a testing program, differs very little between the
two tests. In a high volume test program the per vehicle cost difference is
estimated at 74*; the per vehicle cost for the ASM is about 5 percent less
than for the IM240.
Table 7.2
Coefe Components and Cost per Vehicle for the ASM and XM240
TM24Q
$2.40
$1.75
$1.39
$1.71
$9.12
$16.37
Inspection Staff
State Oversight
Test Equipment
Building Modification/Construction
Other Overhead
Total Cost Per Test
$2.40
$1.75
$0.65
$1.71
$9.12
515.63
113
-------
8 . evaluation of the Adequacy of the ASM for Enhanced I/M
Prograa*
8 .1 Introduction
The preceding chapters show that the four-mod* ASM teat is not equivalent
to the IM240 on a par-car basis. Evan if ASM outpoints are selected so that
the same number of cars are failed, they will represent a smaller portion of
the fleet's excess emissions, and the cars will not be repaired as effectively
as if the IM240 were used for reinspection after repair. However, to some
extent this loss of emission reduction can be compensated for by improving
other I/M program features to make them more stringent than would otherwise be
required to meet the emission reduction performance standard in EPA's rule for
enhanced I/M programs. Among these other features are the inspection of
heavy-duty gasoline-fueled vehicles, the use of the ASM test for all 1981 and
newer vehicles rather than just the 1986 and newer vehicles which are assumed
to be tested with the IM240 in the model I/M program, a higher failure rate
for pre-1981 vehicles, purge testing for more model years than in the model
program, and more comprehensive tampering inspections.
Whether these improvements are enough to offset the loss of benefit
from the ASM is the decisive question that determines whether areaa subject to
the enhanced I/M prograa requirement can rely on ASM testing instead of IM240
testing. Also of interest ia whether it is possible to use the ASM and still
operate a biennial prograa. To answer these questions, EPA examined annual
and biennial scenario* in which the ASM outpoints were made a* stringent as
EPA believe* is consistent with good engineering practice and the possible
.offsetting prograa improvement* were made as large as EPA considers reasonably
possible. If this hypothetical best-possible ASM prograa cannot satisfy the
enhanced I/M performance standard* then no ASM prograa can.
Regarding beat-possible ASM outpoints, EPA ha* assumed that the
failure rate associated with the most stringent IM240 outpoint* for which EPA
ha* provided eaiaaioa reduction credit* is the limit of good engineering
practice in a* I/M prograa. These IM240 HC/CO/NOx outpoints are 0.6/10.0/1.5,
compared to- the> 0.1/20.0/2.0 used in the model enhanced I/M prograa. Th* ASM
outpoint* tha* aatched thi* failure rat* in the full Mesa lane saaple were
0.40/8.0/1.•„ The**) ASM outpoint* can be expected to produce a higher error
of commission rat* than the 0.6/20.0/2.0 IM240 outpoints, but ia the interest
of exploring the limit* of ASM testing, EPA assumed that thi* did not make
thea unacceptable. KPA calculated MOBILES* I/M credit* for the*e ASM
outpoint*, uaing the saae basic approach as originally used foe the IM240
credit*. We thea used MOIILSSa with these credit* and appropriate aaaumptions
for the offsetting prograa improvement* to determine the overall benefit of a
114
-------
beat-possible hypothetical ASM program. Further description of this process
follows.
8 .2 MOBIUSa Analysis
The Z/M credits for the ASM test procedure were determined using the
identification rate from the Arizona test sample. The laboratory sample was
weighted as described in Section 5.2.5.8 to reverse the effect of the
recruitment bias. The fraction of total emissions identified by the ASM test
with best-possible outpoints and the ZM240 test with its standard outpoints
were determined for that sample*. Using the ZM240 results for the Arizona
sample, the ASM identification rates were converted to a fraction of the IM240
results. These fractions were then used in the Z/M credit model to adjust the
ZM240 identification rates used in MOBZLK 5 to represent the effect of the ASM
test.
For repair effects, based on current information, EPA can only give the
ASM teat the same repair effect as the 2500/Zdle test procedure for HC and CO.
For NOx, the ASM test was temporarily assumed to have the same repair effect
as the ZM240 test procedure using a 2.0 NOx outpoint, the nearest available to
the 1.8 ASM outpoint. At this time, we made this temporary assumption for NOx
so that the ASM program can be analyzed for all three pollutants even though
the repeir effectiveness problems found for HC and CO appear to be similar for
NOx. Unlike HC and CO, there is no set of alternative repair effectiveness
numbers available that could be used since steady-state tests have never been
used for NOx control in the past.
Using the ASM credit set described above, we proceeded to perform MOBZLESa
runs for four separate Z/M program scenarios: a no-I/M run, an enhanced Z/M
performance standard run, and two ASM runs, one assuming an annual testing
program, and the) other, a biennial program. All four scenarios assume
national default inputs for the) local area parameter record - including
vehicle registration mix, ambient temperature, average VMT, fuel KVP, average
speed, etc. - and cover evaluation years ranging from 2000 to 2011. Depending
on the osone> classification, statee must show in the 1993 SZP that the Z/M
program selected meat* the) performance standard in these evaluation years.
Both ASM runs) were) identical, with the exception of the. above-noted
difference) in teat frequency. The other program parameters assumed for the
* For convenience in calculations, MOBXUKSa Z/M credits for a particular test
and outpoints are developed starting with the total emissions identification
rate, rather than the) excess emission identification rate used in earlier
sections, to more readily display the) relative effectiveness of tests. The
difference does not affect the) final result.
115
-------
ASM run* include* • program start year of 1983, a test-only network, and ASM
tasting of model year 1981 and latar light-duty vehicles and light-duty
trucks. The ASM runs also assunad evaporative system purge and pressure
testing, and visual inspection of the catalyst, inlet restrictor, gas cap, air
pump, EGR, tailpipe lead test, and PCV system on all 1971 and later model year
vehicles. Full purge benefits were given for ASM testing, since ASM purge
testing will fail virtually all cars that would fail the IM240 test. A pre-
1981 stringency of 40% was assumed, along with a 3% waiver rate and a 96%
program compliance rate.
Once these MOBILES* runs were complete, we compared the results for the
enhanced X/M performance standard run and the ASM runs with the no I/M case to
determine what percent reduction was required to meet the performance standard
and what reductions could be expected from the annual and biennial ASM
programs modeled. The results are shown in Table 8.1.
11C
-------
Table 1.1
MOBXLBS* emission Factors and Reduction* from ASM Testing
2000 No I/M
Enhanced Performance Standard
Maximum Annual ASM
Maximum Biennial ASM
2001 No I/M
Enhanced Performance Standard
Maximum Annual ASM
Maximum Biennial ASM
2003 He Z/M
Enhanced Performance Standard
Maximum Annual ASM
Maximum Biennial ASM
2004 He I/M
Enhanced Performance Standard
Maximum Annual ASM
Maximum Biennial ASM
.2001 No I/M
Enhanced Performance Standard
Maximum Annual ASM
Maximum Biennial ASM
2011 So I/M
Enhanced Perfoxaeme* Standard
Maximum Afuraafc.iam*
. _ «e-
voc
g/m Redux OK?
2.88
1.96 32.0%
2.00 30.5% NO
2.07 27.9% NO
2.66
!.<• 36.6%
1.11 31. i% NO
1.17 29.4% NO
2.52
1.53 39.2%
1.71 32.3% NO
1.7< 30.1% NO
2.47
1.47 40.3%
I. ft 32. <% NO
1.72 30.4% NO
2.3»
1.3* 41.1%
1.60 33.1% NO
1.68 30.9% NO
CO
g/m Redux OK?
22.23
13.98 37.1%
15.08 32.1% NO
15.79 29.0% NO
NOx*
g/m Redux OK?
2.27
1.97 13.5%
1.93 15.0% YES
i.9« 13.9% res
2.10
1.77 15.8%
1.76 1C. 3% YES
1.78 15.0% NO
2.02
1.C7 17.2%
1.C7 17.0% NO
1.70 15.7% NO
1.97
1.62 17.8%
1.63 17.4% NO
1.66 16.1% NO
1.94
1.56 18.8%
1.60 17.6% NO
1.62 16.3% NO
* with t
ry assumption for NOx repair benefits, am described la text.
By comparing the ASM results to the performance standard/ we conclude that
neither the •"H*"**^ annual nor the ITU if"™ biennial ASM program would meet the
performance standard for BC or CO for any of the milestone years. For MO*,
the biennial ASM program with the temporary assumption for NQs repair benefits
meets the performance standard in 2000, but misses it for each successive
117
-------
milestone, while the annual ASM program meets the performance standard through
the 2003 milestone.
These NOx results include a caveat, however. The degree to which the ASM
NOx benefit in the table exceeds the performance standard is quite small, if
the percent NOx repair benefit for ASM testing is anything less than 90%
(i.e., 13.51/15.0%) as good as for IM240 testing, the Maximum Annual ASM
program will not meet the NOx performance standard in 2000. The corresponding
"actual values* for the Maximum Biennial ASM program in 2000 and the Maximum
Annual ASM program in 2003 are 98% (13.51/13.8%) and 97% (15.81/16.2%),
respectively. While EPA for the present reserves judgment on exactly how much
NOx repair benefit is lost with ASM testing, (while we consider a test program
to further explore this question) it is clear from Section 5.3 that the loss
is certainly at least 10%. Thus, ASM testing cannot meet the performance
standard for any pollutant for any milestone date, and therefore is not an
acceptable test in any enhanced X/M program.
118
-------
9. Appendices Table of Contents
Appendix A Teat Procedures A-l
Appendix B Emissions Data Listing for Vehicles Receiving Both Lane and
Laboratory Tests B-l
Appendix C QC Criteria for ASM/IM240 Database C-l
Appendix 0 IM240 Cutpoint Tables D-l
Appendix E ASM Cutpoint Tables E-l
Appendix F Scatter Plots and Regression Tables F-l
Appendix Q ARCO, Sierra, Environment Canada Data Analysis G-l
Appendix a Estimated Coat of High-Tech I/M Testing H-l
Appendix X ASM and IM240 Credits for State Implementation Plans With
MOBILES Runs 1-1
Appendix J Emission* Analyser Price Information from Horiba J-l
Appendix K Centrifugal Blower Price Quotation from. Combined Fluid
Product* Company K-l
Appendix l> fae*-Paa» and Fast-Fail L-l
119
-------
Appendix A: Test Procedures
-------
Test Procedure to Evaluate the Acceleration Simulation Mode and
the Emissions Measurement Capabilities of a BAR90 Certified
Analyzer With An Integrated Fuel Cell Type NO Analyzer
1.0 Objectives
The objective of this ASM test project was to collect data to compare the
effectiveness of a four-mode steady-state test procedure as an alternative I/M
test to the ZM240. Emissions and canister purge flow data were collected
using the following vehicle operating modes:
- Two Acceleration Simulation Modes (5015 and 2525)
- A 50 mph steady state mode at road load
- An idle mode in Drive
- An idle mode in Neutral
These modes will subsequently be referred to as the ASM test. The same
data were collected for the IM240 test.
The lower cost of the emissions measurement equipment is the salient
feature that makes the ASM attractive to its proponents. Therefore, EPA made
ASM emissions measurements using a certified BAR90 analyzer (for HC, CO, and
C02) with an integrated NO analyzer of the fuel cell type for NO measurements.
For the IM240 a CVS-based emissions measurement system was used.
Canister purge flow measurements were made with the same 0-50 liter per
minute for both the ASM and ZM240.
The testing was carried out in two locations, a single I/M lane in an
official Arizona I/M station and at a laboratory owned by Automotive Testing
Laboratories (ATI.). Both were located in Mesa, Arizona.
2.0 Phoenix Lan« Procedure
The following is a description of the I/M lane procedures.
- This procedure was restricted to 1983 and newer light duty vehicles
with fuel injection, when available. Carbureted 1983 £ newer
vehicles were tested when fuel injected vehicles were unavailable.
A-2
-------
Pre-1983 light duty vehicles were tested only when 1983 and newer
vehicles were unavailable.
- Each light duty vehicle received:
1. The ASM test that included the following modes in the sequence
listed:
- ASM5015 with purge,
- ASM2525 with purge,
- 50 mph at road load, with purge,
- idle test (automatic transmissions in drive),
- idle test (automatic transmissions in neutral) for the first SO
cars. Car 51 and subsequent cars will not get the 5th mode.
These four or five modes will be referred to as the ASM series.
2. An IM240 with purge.
3. A pressure test.
4. An Arizona State I/M test.
2.1 Procedure Sequence
- In general all odd numbered vehicles got the IM240 as the initial
test and all even numbered vehicles got the ASM as the initial test.
- Data collected included a number 1 or 2 in a field named "Test.Order"
to designate whether the ASM series procedure was run first or
second. Discrepancies between the Test.Order entry and even/odd
vehicle numbers are resolved by relying on the Test.Order entry, as
this was ATI/s primary means to identify test order.
2.2 Measurement Equipment
- For the ASM series, a certified BAR90 HC/CO/COa exhaust emission
analyzer was used to measure HC, CO, and CO2, with an integrated NO
analyzer using a fuel cell sensor. ATL only acquired a NO
analyzer/BAR90 analyzer combination that provided second-by-second
data for HC, CO, C02, and NO. The data output for the ASM test went
A-3
-------
to 3-1/2 inch floppy discs that included run number, time (sec), mode
number, vehicle speed, purge flow, NO (ppm), HC (ppra), C02 (%), CO
(%), actual torque, required torque, actual horsepower and required
horsepower.
- A 50 liter/min Sierra flow meter was used to measure total canister
purge flow. The flow meter system output was the cumulative second-
by-second data for total flow recorded on the 3-1/2" floppy discs
discussed above.
- For the ZM240, normal measurements with the CVS system continued at
the lane. The data collected included time (sec), bag number,
ambient measurements, NOx (grama/second), HC (g/sec), C02 (g/sec), CO
(g/sec), and purge in standard liters.
2.3 Procedure Details
- An electric Clayton dynamometer was used for both the IM240 and the
ASM series. The dynamometer horsepower settings for the ASMS were as
follows:
• 5015 HP - (ETW / 250)
• 2525 HP - (ETH / 300)
• 50 Mph HP - Road Load
The horsepower and inertia weight settings for the ZM240 were as
normally performed. The minimum inertia weight setting (2,000 Ibs.)
was used for the ASMS.
- Manual transmission vehicles were tested in second gear for both the
ASM5015 and th* ASM2525. The 50 mph road load mode used the top non-
overdrive gear, typically 4th gear on a 5-speed, 4th gear on a 4-
speed, and 3rd gear on a 3-speed. Drivers used a lower gears for
vehicles that were lugging.
- The engine was shut off prior to the IM240 and the ASM5015 (as will
normally b« don* by I/M programs to connect the purge meter),
regardless of which procedure was performed first, and restarted just
prior to initiating these procedures. The engine was not shut off
between ASM modes, and the vehicle was accelerated from the current
mods up to the next mod* speed, without first returning to zero.
- The ASM emission sampling period and the canister purge flow
measurement period were as follows:
A-4
-------
1. Each ASM mode was initiated after the vehicle speed had achieved
the nominal speed (15, 25, or 50 mph, and 0 mph idle) ±2 mph. Once
up to speed, emissions sampling of one second average concentrations
continued for 40 seconds. Emission scores for HC, CO, C02 and NO
were reported for each second. Emissions scores for the first 10
seconds of each mode were ignored to allow the dynamometer to
stabilize and to allow for transport time to the analyzers.
2. The purge flow reported was the second by second cumulative flow
over the entire ASM cycle, including transient accelerations. The
nominal acceleration rate was 3.3 mph/sec., with a minimum
acceleration rate of 1.8 mph/sec and a maximum of 4.3 mph/sec. The
table below lists the minimum, nominal, and maximum acceleration
times used to accelerate from one mode to another. For example, the
table shows that the time to accelerate from 25 mph to 50 mph should
be 7.6 seconds., but can take as long as 13.9 seconds., and as little
as 5.8 seconds. The zero to 60 mph time is provided to indicate how
the specified acceleration times relate to a commonly known reference
of vehicle performance. ATL used a video driver's aid with the
nominal acceleration rat*.
Minimum
Nominal
Maximum
Acceleration
Rat*
(mph/sec)
4.3
3.3
1.8
Time
0-15
mph
(sees)
3.5
4.5
8.3
to Accelerate from- to:
15-25
mph
(sees)
2.3
3.0
5.6
25-50
mph
(sees)
5.8
7.6
13.9
0-60
mph
(sees)
14.0
18.2
33.3
During the accelerations between modes, the dynamometer load setting
did not exceed road load. This was specified to enhance the
opportunity for canister purge during the ASM accelerations. The
combination of the ASM load and the base 2,000 Ib. inertia may load
some, vehicles to heavily to allow purge to initiate.
3.0 Lab Reeruitmant
Light duty vehicles that received all of the lane tests (IM240, ASM
series, and Arizona I/M test), were recruited for testing at ATL's laboratory.
Cars were categorized as passing or failing using the IM240 cutpoints in the
table below:
A-5
-------
Phoenix Lane IM240 Outpoints for Lab Procurement
Model
Years
1983+
HC
0.80
CO
q/mile
>15.0
NOx
q/mile
>2.0
The following table provides the laboratory recruitment goals for the
pass/fail categories listed as a percentage of the total number of cars
recruited to the lab for this task. The initial recruitment target was 100
vehicles.
Phoenix Lab Recruitment Goals Using Lane ZM240 Categories
Model
Years
1986+
1983-85
HC/CO
Pass
15%
10%
HC/CO
Fail
15%
10%
NOx
Pass
15%
10%
NOx
Fail
15%
10%
4.0 Commercial Repair Recruitment
Owners of vehicles that failed the Arizona I/M teat, and received and
IM240/ASM series, were offered 950 to return to the lane for after-repair
tests. These vehicle owners were only offered this incentive if they refused
to participate in the laboratory testing program or if their vehicles were not
needed for laboratory recruitment. Recruiting vehicles for laboratory tests
was a higher priority than for commercial repair participation.
The owners were informed that they must return with repair receipts
indicating repairs by a commercial establishment with itemized labor and parts
costs to qualify for the $50 incentive. ATL included either the original
receipts or copies in the vehicle test packets that ATL provided to EPA. In
addition, ATL provided summarized comments and data for these vehicles on
electronic disk.
Vehicles returning after commercial repairs followed the same procedures.
5.0 Lab Procedure
The lab procedure is summarized in Attachment 1, so this section will only
add explanations to the procedure listed in Attachment 1.
A-6
-------
5.1 Two Group*
The vehicles recruited to the lab were separated into two groups:
1. Those whose initial lane test was the IM240 and were repaired to IM240
targets. For the vehicles in this group, the ZM240 always precedes the
ASM series (see Attachment 1).
2. Those whose initial lane test was the ASM series and were repaired to
ASM targets. There were not enough data to set ASM repair targets, so
IM240 targets were used for both groups. For the vehicles in this group,
however, the ASM series always preceded the IM240 (see Attachment 1).
5.2 Repair Target*
The repair targets were to achieve 0.80/15.0/2.0 on the IM240 for both the
ASM-targeted group and the IM240-targeted group. Initially, repair targets
were to be provided to ATL for the ASM targeted group to replace the IM240
targets. However, due to time and data constraints this proved impossible.
For the initial repair attempt, the mechanic was only aware of the lane
IM240 score for both vehicle groups (initial lane test: ASM or IM240). For
subsequent repair attempts, the mechanics were only be aware of lane and lab
IM240 scores. FTP scores were not provided to the mechanics for either group.
Repairs were limited to 91,000.
5.3 Laboratory Te«t Equipment
Due to time and financial constraints, EPA was unable to develop lab ASM
capability. Th« IM240 and FTP were measured with a CVS system. Modal or
second-by-second CVS capability waa not available at the laboratory.
A-7
-------
Appendix A: Attachment 1
ASM/IM240 Lab Procedure
Revision Date: 10/21/92
Number tested *
Recruitment: 1983+fuel injected only.
Repairs: Get IM240 Indolene to
.8/15/2.0. The mechanic should only be
aware of IM240 scores for the IM240
targetted repairs. $1,000 repair limit/car
- catalyst if necessary, aftmrkt preferred.
Develop explanations for any IM240
failures that pass FTP, while veh is still at
lab.
Tank Fuel
On-Road Warmup
Tank Fuel IM240
9.0 RVP Indokoe As-Received
LA-4 Prep cycle @80°F
No Diurnal
FTP Exhaust
No Hot Soak
IM240 Indolene (with purge if available)
Repair to get IM240 Indolene to
.8/15/2.0, The mechanic should only be
aware of IM240 scores - not FTP scores,
•UI0 QQiy DGTZQRB IDilltllZUIXl FCDftlTS
necessary to achieve targets.
Report After-First-Repair Indolene
IM240 regardless of outcome. Mechanic
will only be aware of lane IM240 score
for first repair, not lab tank fuel score.
Continue repairs if necessary. Dont
perform FTP until .8/15/2.0 is achieved.
9A RVP Indolent After-Repair to
IM240 01/15/10
3 LA-4 Prep cycles® 80°F for all
vehicles
Top off to 40% fill-don't drain.
FTP Exhaust
IM2U tatatene RM1 (w/purge if
available)
Stop repairs even if failing FTP.
Indolent Lane Tests For Vehkkf
Whojt Initial LSUM Test Wss IM240
Indolenc Ltnt Testt Procedure for
VeUdes Whow Initial Lant Test Was
ASM Serial
On-Road Warmup
Lane Indolene IM240
ASM Series
On-Road Warmup
ASM Series
Lane Indolene IM240
-------
Appendix B
Data Listings
-------
w
0)
o
-,-t
x:
o
0)
an
CO
co
•H
(0
s
0)
3
(TJ
EH
.U
c
•H
O
a
o
53
8
%
OOOCMOi-IOt-IC4Oi-«i-IC4<-l<-IO>-li-IOOCMOOMOOCMi-«i-ICMOOO
,-ir-i««-icMoartint~f-ii-ir>
OOiHOOOOO«-IOOOOOOr-«OOOOOOOCS|OOi--liHOr4i-IOOO
8
. . «o r-
(N
I
03
3
c<it-ii*»eiic«i«*c
oddodoooooooooooooooooooooooooooo
o>r»
Oi-«OHr4r4i-lf-l^r4f-ltHr4<^i^f-«
-------
CO
0)
r*
o
•H
£
3
TJ
CU
j-i
3
mONCM<*»vor»cor»o»oocot-if-4f*r»r-^<r><»>>nvofM0»o*o*a
f-tcNNvvOi-tveoom
O«*OOO>HOf-1
<-\
<0
5
0)
l-l
X3
m
H
•» « u, CM W.
« rH ' CO °*
CM
1
S
a
en
i
CO
2
cMO»tHO»^t»a»p«co«»»1_vo»cort(o
inCOCOM'*'
CO CM O» CM
Aj*j'Oij'OT3tJ>OT}'O^4JW4J4J-a-OAJ4jjo-ouT3'O'Ofc»*jjO4j'a'aiOw
ddodoooooaoooooeoooooodoeeeoeoooo
§
((MCMCMCMf^(MCMCMCMNCMCOCOm(0«*>
-------
w
0)
iH
o
•H
ff,
0)
T3
oic>if*»nf><»>o>io>ir-i<-i»«-i<-iinin»-i.-4evj
dcsioodooodoooooooooooooooooooooooo
01
8
(0
>
rH
(0
5
(I)
rH
£1
(0
£-•
\
C
I
M
94
OOOOOOOOOOO°f-«OOOOOOOO
i/) l/> P» U"> U") CNJ ri>r'r«coa>ooaoeo>4r4r4C4(N
rHfHi-li-«<-4*4<-liHr4<-l«-lr-lr4i-li-l^rHr4f>lf-t(^r4i-«fHr«4f-lf-«MCi|O
fHf-lr4i-tr4r4i-lf-lr4r4<-t>H<-lf-l<-l<-4r>oo«o«o«A9>o\o<-«(M«0«mvmioininin«vaaot0iOf<4(N
!
Mo«oiPfo«n«
in 10 in m in i/)
-------
3
I
Bi
i
a
o
'
i
1
II
8
4
•T
|
1
I
I
I
!
i
i
i
i
0
j
1
1
1
(
o
2
o
(
1
CM
I
9
t
;
|
£
g
s
5
•
8
c
c
|l
y
:
•
I
*
*
C
?
s
£
2
g
W
a
9
jj
*
8
y
V
j*
1
3
t
|j
2
4*
5
f*
4j
S
3
*
41
CM in envo vo r- o vo * M VOM cocn MOO OM co * MM <-ico
vo cn MVO * eo m vo OM in vo coco cocn in o vo o m * co<-i
OO COO MM OO MM MO OO MM i-IM OM OO rHiH
* * voen co» o en r* m *"! * «*** * eo voo '"") r~ COM o M
J^co COM r- m vo* coen 2 *"* co "^ covo men r?"1 coco cn co
in co voen * en * M mm men eoen coen enen r*vo r-ien ooo
COM coo MM cncM MM enM * M MM coo m M MO CMO
MO OO OO OO OO OO OO OO OO OO OO OO
^^ ^^ ^* f^ QO ^^ i*4 ^* ^* ^0 ^0 ^^ ^ft *^~ ^^ GO ^* ^J*
•M * * ^^ * * f^ * * * f\ * rt\ * * * * * «M * * * irt *
^^ ^^ ^^ ^^ ^* f^ ^^ ^^
cn r-i o vo co co vo r- m M vom mo com oen en co cnm MOO
r-M MM MM oen *M MO COM MM mo r»M oo voo
OO OO OO MO OO MO OO OO MO OO OO OO
mo Mr* * en o m r— vo moo oo* vor* covo r*>en o* or*
en* in oo CM m CM r* OM enM *M CM co *en cn* cn* men
OM *O CM M MO MM MO OO CMM MO OM COO MM
« oo o * * r- 1 « M « «-. cn « vo eo * cn * ^ m M oo « vo
P{M Mr* ** <5* vo* J^1^ ** MM r»o *o MO ***
*en o* *M r-o m«o ** *M MVO oo oo mr» en* vo*
*M MM co en r- cn * M coo C*>M MM mo COM MO r*<-t
MO OO OO CMO OO MO OO OO MO OO OO OO
en m MOO *m vo r» r» M * en MCO voen ooo VOM co M oovo
men co eo OM M en OM M m COM M* men en 00 r- m coo
OO mo MM OO MM MO OO MM MO OM OO MM
co m * n « cc
MM MCM MM MM M)W MM MM MM M«M MM MM MCM
OO OO OO OO OO OO OO OO OO OO OO OO
22 22 22 22 22 22 22 22 22 22 22 22
MM MM MM MM MM MM MM MM MM MM MM Mfi
com com om 0*0 en m 000 o co om mm r»m vo en eo*
MM MM m M mo MM en M enM oo M OM MM oo CMM
OM OM OM O M OM OM OM OM MM OM MM OM
MM MM MM MM MM MM MM MM MM MM MM MM
n cn cn en cn cn cn en cn. cn on 01 on cn cn en on on OK on o^ o^ en c
5 {/j *"J *jj •jj t/\ *2 co S co T! Co c/i co *^ co co ^« c/
14 m. CO CQ w] £^ Vt A? CO ZZ CO 22 ^O K^ ^O c« ^O ^O CO £^ CO K^ vi «
H^ HH H*^ H"^ H*^ H"^ H^ H^ HH H^ H*^ H^
M» vom cnm mM vo vo r*« MM enm OM r*r» min oo*
C*»M on* mM mo MM »M mm *M M* P«M onm oor-
vo cn VOM r* cn r- tf» r» cn r* o r* o r- en oo M vo on com vo ^
MM MM MM MM MM MM MM MM MM MM MM MCM
OO MM »» MOT vow c»t» encn oo mio cnon MM *•*
mm mm mm mm mm mm mm vovo vo vo vo so r» r* r* r-
MM MM MM MM MM MM MM MM MM MM MM M <->
mm mm mm mm mm mm men com mm mm mm mm
i
ca
-------
CM
t
i
m
I
co CM vo vo vo m
o CM * c* in m oo
fH rH rH O •-* t-l
vo o
vo
^ co
in
vo ro OD vo v
CM CM » <-i m oo
O 0 0 O 0 O •»
O O r-t
in oo m
m oo oo
en vo m
vo CM vo
r-l O O O oo o
O O O t-4 rt CM
-»
« a
S S
« -
s
S S
o o
•O fl
nee
0 0 0
M e «
o o o
cn vo
O fH
CM
CM CM CM CM CM CM CM
O O O O O O O
CO CO CO CO CO CO
en en en en en en en
H K H fr« H (t frt
n o> M co co o) co
* K H H H H H
o oo in o en o
in co r* eo t-« co
CM CM » * » m m
CO CO CO CO CO CO
r« en o ,-» CM » m
in m vo vo vo vo vo
M CM CM CM CM CM CM
O CO CO CO CO CO CO
in
i
en
-------
gj
I
XM240
4
w
4
|
• Scores
3
•
i
<
u
«
M
(
1
M
I
a
•
4.
•r
a
i
1
•
8
2
«
3
2
M
g
s
ft
r;
a
j
a
in
|
X
8
a
u
1
0
j
Q
tt
I
1
»r * oro eo » or» vo " * ™ <* oo M v vo en en t-* CM M
™ * 2 m com £ m *<*» M ^ 2 ^ ^ <*> *- * N ™ "* * rr)'r
p- co oom in vo MO M vo o en <-i vo Oi-i vo M mm p> r- co M
rH O «T |H r-im r-( rH VO M t- coo r-i«o r»co CM i-t o m
o» c»o om OM cMt-i p-rH ini-i coo mo c\im inm CMCM
I-IO rH O rH O rH O CO O IT) O (NO OO rH O OO OO OO
mo or* v (N vnm o ve on vo ov vo r- i- m r» o CM i-i i-i«o t- vo rn m voo oo vco coin covo
OO rH rH CM rH O rH OO OO irt O MO MO MO rH rH r- rH
•vo °!^ incB °. r^ * r* • r» ^rHmmmco^m coco ^M
5°* In0* co» IN" S01 <5N rHrt «oo coo i-»o com mr-
om * p- m P- enr* Mm voo co vo eno I-H m cn« inert mm
O>*> COO mm M rH O rH O rH r- rH O P- O MO OO rH O OO rH O OO
VO rH VOOO M » O M coo vo » r- rH in CM
OO M rH rH O Ov4 OO OO lOO MO rH O OO rH rH P-rH
•^ en ^,4 co oo "Jen ®r4 °. w "iMinr-^enrnvo f.,H ^rn
J3 p« p» o»« Jg«n ^in 5® ^r^m^o^o^ S* ^^
OM OM voo> mm on m n0* P>V en« p- co eft M «r> OM
«« Oft M VO OT» » in » o P>** r- co in eo ao o p- p» r-trn
OO rH M rH O rH O MO MM CM rH MO MO M rH MO O rH
rH M rH M rH M MM MM MM MM MM MM MM MM MM
MM MM MM MM MM MM MM MM MM MM MM MM
en en en en en en en en en en on en cnen encn en en en en o\ en cnen
MIT M£! MM M vj MM MM MM MM MM MM MM MM
Hn ** 55 H H H 5 ** ** H ** H H ^ ^ ^ fr* *•*• t4*4 H *^
w M w ZJ cn o» co 2 coca coca coco coco coco coca coco coco
P. *i vo vo mm * « OM M vo OM COM COM mr> p- co mr»
o vo voo o m MOB vo ao vo m COM mm en m «m moo p>co
en«*> enM o* OM M M»
MM MM MM MM MM MM CMM MM MM MM MM MM
mm cnen oo mm coco oo vovo r- r- co co oo MM MM
r»r» r»r» co co eo eo coca en en cnct cnen en en oo oo oo
MM MM MM MM MM MM MM MM MM MM MM MM
mm m m m m m m mm mm mm mm mm mm mm m<*
^o
i
CQ
-------
(0
m
a
£
flj
a
0)
os
o
M
, t
4J
c
o
CJ
*
M
,
•3
8
I
M
O
i
«
u
1
0i
B
I
^
H
3
3
i
i
c
t
w
*
1
c
u
c
c
a
N
C
H
|
0
a
I
0
{
*
1
8
u
4
3
i
§
a
a
u
1
2
U
g
«
8
y
M
M
41
0
t
^
o
41
'
m
I
*n
1
cMen * fi COCM p- co mm * m mvo mft Ofi o oo eoen men
m * oco p- vo i-tm vo* oen mvo om co p. p- * CMCM ocr\
m ft r-ift oo mo oo *»• o oo CMO r-tfi ft o mo mo
•«• en * * ^ °*» vom ^ * P- co oen vOft ft OCM * fl
OO r-l O r-IO OO OO OO OO OO OO OO t-tO OO
^ 22 "2 ^0 IMO 22^-'op^°*o £° *"
r-i CM enm CON oo *N mm en CM men *en mm mm voco
i—co ftco coo en ft eno i-ifi mft coo meo eno en o oo
oo mo ft o oo fto oo oo fto oo oo oo f-to
P-N p-m fift P-N coco mm p-m *en CM* co* p- vo five
mm en m *vo en ft mco ft o> 1-1 o enm CM m p- * TO o vo
*fl Oft OO NO OO mo flfl CMO CMfl ft O *O *O
^cnmenf
• t ^^ • • ^^ t • • ^^ • ^^ • « ^^ ^^ • • ^^ •
*4 P* P* fi en ^^ ft CM ft fi ^^ fi ft
*p> oen VON CON N* m* *co o ft men mco m* *co
p*eo NN ooo Oft *o fifi m ft o ft mp* oo ft o oo
OO COO ft O ftO ftO OO OO NO OO ft O fl O flO
p»o CMfi ft vo ft* p*o ft m mco ON CM vo vop* cove *m
ftm en ft men CM m mm ft o enp* mm oco mm CM m com
*fl O ft OO ftO OO flfl OO CMO CM ft fl O COO r-IO
^
~ p* ^ r^ r^ , %M t ^9 w» p» rn ** % \o p* ^ t 'W t ^F *o \o
* * f* * CO • ^j • f^ • • • • • _^ • • • _.. • ^ • • •
^^
vom m* GO N o* coco p-m mo mm or* N* oo mm
en p* co * co 01 * co * co co co co co fi fi co vo N c^i co CM P* ft
OO COO .NO ftO OO OO OO ft O OO ftO ft O OO
aa aa a c an c e ce c c aa c a ee aa aa
flfl flfl NN f»t* NN NN CMN flfl NN NN flfl flft
OO OO OO OO OO OO OO OO OO OO OO OO
SIS! SISI 3 $ «I9 9>£! 51 51 9*1 «l«l 53 9 99 99 99
op> cno m ft CON coo *N m CM m* m * eo r» en* fifi
m fi o ft o ft ON o ft o ft Oft Oft o ft Oft ON ft CM
ft CM NN NN NN NN NN NN NN NN NN NN NN
NN NN NN NN NN NN NN NN NN NN NN NN
fl £1 fl N fl N fl £1 fl N fl N ft CM ft N ft N fl £1 <-l|* *••**
H R HH HH H R HH HH HH HH HH N SJ H£ H£
w rj COM co to car* con coca coco coco con co r* wzf COL
H^ HH HH H ^ HH HH HH HH HH H*^ H^ H*^
r-t en en en * eo p-m o>-t N* com en ao vo m *o fto meo
vo vo fi m CM vo co* co^o mot voen vo ft p- ft NP» * eo coo
Ntn co* co* co«o co* co* m* mm coto *io *ve *vo
NN NN NN NN NN NN CMN CMN NN NN NN NCM
cn
-------
n
i
XM240
•
u
I
a
£
•
!
I
4
i
i
1
i
(
0
:•
41
•f
n
<
c.
«
u
(
a
N
t
a
•
4-
•4
•
!
i
\
»
s
i
»
%
n
«
s
B
|
z
a
s
z
s
M
|
2
2
*1
2
«
1
r- T mo a>o co "T en CM en CM vo vo m o p» vo en vo en CM n IH
o co m co en » m* co m en *• vo co o m en en oo vo m o co
r-tO » v«n
oa> ^o r» moo H<-I CM CM oo 1-1 <*>m
OO OO OO OO OO OO OOOO OO r-t O OO r-l O
o\o CM oo •» rH oo CM i- co (M<-i oo m ro o *-tcn m p- vor> mi-i
«"O\ m»-« f>m cn» CM oo
oo o f-t Oi-i oo CM o>-« «o vo o CM o r-t t-t m CM en i-i en«>
OO OO r-IO OO OO OO OOOO OO -1 O OO r^oo oop» P»CM com cnmi-ico men vo o *p» ovo
p»eo ft CM en co •-• CM * » vocn cnromco CM vo o CM co i-« en CM
OO r-l ft <-IO CMO OO r^ O r-IOi-IO CMO f-lfl <-IO OO
mo p- vo oo« m oo ^-i eo p- vo * oo co m co vo rH m "!<-• vocn
i-ti-i CM*M m*^ «orM «n« »o ^O«O»H co^ mt-i JJ^ 00*0
TO p»«o en m r»r4 CM m vo<-i T CM en p» oca moo co CM voo
CM co CM CM CM W *J T3 -O MM T3 T3 -O T3 4J aJ 4J 4J 4J *» W AJ
ec c c a« n« ee «n c c c c « « n n n n n n
oo oo oo o o oo oo oooo oo oe oo oo
22 22 22 22 22 22 22 22 22 22 22 22
MM MM MM MM MM MM MM MM MM MM MM MM
M *«n m» VO«VOM p»«» r«r» «or» com
r-l CM i-tfM MM M(M <-• CM M(M MCMOCM OM OM M(M CMCM
CMCM CMCM CMCM CMCM CMCM CMCM CM CM CO CO CO <*> CO «n CO CO COCO
en en en en en en en en en en en en cnencncn en en en en en en encn
^2 ^£1 "^£1 «^£! ^2J ••«£? M£*MIM MCM ^CM «-»fM <-«M
"B "S -s -£ ^s "^ »g «^ «« «s -^ ^^
H*^ K1* H" H** H •* H** K^HK HH HH HH fr»H
o*« CM«O encn p» o *OM en en co -i o m mmmwrncMcor*
en» enp> en vo or* MC* cMr> »oocoo o «<0 m «o m *o m vo mvoooM coo aai-« en •-» CM«O
CMCM CMCM CMCM CMCM CMCM CMCM CMCMCMCO CMCO CMO> CMCO COCO
encn oo CMCM com * » mm r« p- co eo en en oo CMCM P» P-
CMCM CMCM CMCM CMCM CMCM CMCM CMCMCMCM CMCM CMCM CMCM CMCM
cof> row «•>«•> «*>«•> coco coco cocoroco coco coco coco co
-------
2!
2
8
(M 10
(M en
vo m
p. «o
m o
m M
O O i-lfl
«*> M
o o
oo
o» co
C3 O
4* VO
00 M
CM O
in
M O
• •
r-( O
m o
r-l O
OO
r-IO
MO
oo m
co en
oo
oo en
vo p.
• »
o o
VO «O
rl O
«-l P»
CO O
• •
co o
Ot-i
r-l 10
o* o
VO CO
r-l CO
M O
en rH
*• o
M o
in p-
• •
OO
2"
i-lO
p» en
co p»
vo m
o o
OO f-lt-l
M o
• •
o o
8
vo en
• •
~4 O
m
•
M
O O
^^ ^H
• •
M O
•o -o
G e
M M
O O
« II
O O
a «
r4 fl
d d
??
M (M
O O
O rH
M M
O O
cn en
m
S
o oo
m<-i
CM m
en cn
m in
M M
co co
M N
•-• N
M W
O O
f-l M
M M
O O
en 01
r» o
t-t f-i
M CO
O O
en 01
<-4 M r4
co
vo vo
MM
CO CO
co co
MM
CO CO
H
COM
HH
co co
m m
M M
co co
I
03
-------
a
3
*
a
J
c
ti
e
M
4i
a
i*
3£
H
J
*
•r!
j
4
1
j
i
s
'
I
i
»
i
c
ia
4
•;
1
!
c
i
«
0
M
Oi
1
a
j
]
c
J
1
*
^
d
M
Jl
jf
j
1
S
J
1
1
1
t
(
c,
_
i
t
S
j
»
„
c
s
9
g
2
8
3
8
ij
o
9
M
•
5
0
,
3
j
B
^
»
VOOP- P- «-i mo on o rH co en p. ooenen om CMP- CMP* om
m vo o rHp- oo rHvom vo*m rH m co vo m voon oen CM c~
OOO CM rH OO CM rH rH OOO rHrHO OO OO rH O CMCM
•CM* rHcn °. o p- m * °*. °. T * ^ «n "! T ^ CM m. m. * rH
* ' " * ^ la * * * _* .^ «• ift t* * *^ O * ** • *
^4 ^^
VO CM rH *rH VOm *OrH *CM* VOrHCM CM in VO rH *rH P- VO
OOO OO rHO OOO rHrHO rHrHO rHO OO (MCM OO
oo0 00 go ooo ooo ooo oo °,0 oo o0
. T, os rHoi • *"J moo ... ."1*^ ^^ • oo ^T "I1"1
|g}£p-' CM'M* *§{ on** SeoS VOCMO mcM ^^ OCM 2®
rH rH
00 rH rH O rH Ul VO CM VO rH 00 CJV Cft *MCM CM rH OO* O Vrt rHCM
MP>rH OM P»O OrHOn VO rH VO mOP> OOn m rH rH CO OrH
• •• •• •• ••• ••• ••• •• • * •• • «
rH O O OO CM rH rHrHO CM CM rH fO CM rH CM O Irt O m* rHrH
oo ro CM moo rHO vo en rH o en r- rHooo oo vo CM vo en CM I-HOD
* VO * *P- rHM OOinO mO* *OrH CMO P«P» rHM P»OO
OOO CM rH OO MMO OOO rH CM rH OO OrH rHrH OO
oonon vo p- on rH I-H ao rH o on m *vo* co o vo m oen vo o
o m o OM VOM moo * o m vooi-t o en CON o vo *CM
*•• •• •• ••• ••• ••• .• •• •• «•
rH O O OO CM rH rH rH rH CM N rH OMN NO l/> O VO m rH rH
floon rHN rH oo p» vo o rH * onvoon rH ono o*
rH O o rHO ° Q OOO OM* NON *o ° O (-00 OrH
OO OO O OOO C* f~ «-t NOO P> rH O OO
lA ^5 IA lA ^5 IA INI r\i ^ n- O IM «^
**^ m rH **^* OOS* 5*2 rHrHrH 3^« *2 ^«J N*
*^
mr* MM oo m o ao rH m o*m om M cove vo p.
rH On ._ OO * f3 m VO O rH O rH m m rH OOn P- _, m rH *O
• • ^^ • • • ^^ • • • • ft * . . • • • • ^~ • • •
NN OO VO rH O O O O MOO *O 00 rHN rHO
oncnorn o r-r-co
T3^^ ^ ^ T3 "0 «U^ on o r* on o p- on o r~ on r* on p» on p> en
rHrHN ft ft ft ft rH rH N rH rH N rHrHN rHrH ft ft ft ft rHrH
HHH H H H H H H H H H H H H H H H H H H H H H
CO CO CO COCO V» (0 CO CO CO CO W 10 CO CO CO COM COM (OM COCO
00 VO v 000000 OO OO 00 OO rHrH
on en on oo * * p» r» r» on on on en en en mm vo vo vo vo eo eo
ao ao oo vo vo vo vo P. r» r» r»r»f P-P»P» ao ao oo oo onon onen
rHrHrH NM MM MMM MMM NNN NM MM MM NCM
rH rH rH rHrH rH rH rH rH rH ft ft ft rH rH rH rHrH rHrH rHrH rHrH
0
1
CO
-------
(TJ
(TJ
Q
•H
m
a.
r-t
•H
O
M
e
g
o
(j
^
M
2
9
a
8
o
3
0
*
1
u
•
"
Jg
^
H
M
H
it
|
1
3
[
I
•
c
(1
a
4
•r
a
1
C
2
c
£
CM
e
j
0
4J
3
j
3
•
a
*
«
3
M
1
12
*n
3
i
£
•
§
il
p
o
1
fj
1
4
N
s
rj
*
£
SB
8
0
a
8
a
8
,i
*
M
3
0
u
2
•
1
^
I
«•
mm i« m I-ICM * vo CMVO <-ivnoo VOCM en oo r» CM mm vm
moo of cMoo voen mi-i 01-10 IH r> menr>m * in en *r
OO rH CM r-IO r-l O mm r-l iH l-l CO r-l OOOO OO OO
om COCM mm ^ CM m* °°. o CM T oo
rH en CM CM v CM o en vo in in o r* m vo en rH ^ oo CM CM rH rH
OO OO OO i-l O OO r-l i-l r-l r-l O •-! r-l O iH OO OO
00 go °,S m S °, g oSS 0° ".*.«« 00 §0
•• CM* "O VOfH ** O^OCO VO* **f-'{J^ '* p«!*
^ iH __J CM CjV _A -— _. CO VO ,. , . -.. . PO ^* CM w» CT\ ^*4 O !» ^"
^ r^ ^ ^* CM r^ CN " . . CM ^ ***
cneft ^^> ro«-» p*^ 01 GO CM o in cnv CM m co v r*cM CMP-
r-* o ^*f*4 eneo co vo CM o ^* CM co COCM CMCM^*^* oo CM •-*
• * •• •* • • •• ••• •• •••• •• ••
r-io oo oo CM o r-i r-i in m m CM r-i m * r-i r> oo oo
oo CM v p» vo rH en om oomen mo o m m CM co + rnm
CMVO in v p» CM meo mm vo vo vo CM i1 CM CM r-i •-! r~ oo CM m
OO O r-l r-l r-l rH O mm r-l r-l rH * CM O O r-l O OO O r-t
r^N 3 c4 2« S2 S^ ciicQ 2^ S^S^ ^2 Jjjer!
VOCM r-i en vo oo mo mi-t CM •* r> CM » oo m •« oo enen «* c~
rH rH * 1-1 OO CMCO en rn; o r> oo » en eoeoeov oo CM rH
• • •• •• •• •• ••• •• •••• •• ••
•HO OO rH r^ m O rH rH VOmm m rH m « rH OO OO OO
o^ 00 °.o ^o ovo »r»o com CM CM m eo ooo oo
o * o o ocMOOrHommomcM
r-i ° r~ CM ^ vo * vo CMOO en «H ^ °° m mm^m CM
CM CM lH •» rH i-l VO (- VO O\ - *f in VO OO P- Cn rH OO rH rH
o o vo m v en tH enen cn^cM * CM vo en eo m ^ o voo
O rH O OO i-l O OO O°tH OO OO VO iH en O iHO
men ^ ° * '"o °*N ^ en r^*t-»'^ *° OCM
ce ee «« oe cm m m m c c m m « « me mm
CMC* CMOi r-lrt CM CM CM«H H»Hi-l CMCM rH^fH^ t-»CM i-lr-l
* • •• • • •• • • • * • • • • • • • * * ••
OO OO OO OO OO OOO OO OOOO OO OO
22 gg gg 2g gg ggg 33 gggg 33 33
r^ F"T p^ ^^ i^ ^^ ^^ p1^ f^ ^^ r^ r^ r^ p*f p^ ^^ ^^ P^ ^^ p^ ^^ ^^ ^^
OP^ CM r* CMCM) *0** en en mm rH ri
oo oo CMCM CMCM * v oo vo vo vo oo «««« r* r» r» c-
oo <-i rH rH i-i tH fi CMCM CM CM CM mm mmmm enen vv
m en en en en en en en en en en en m en
r-tfH) r-t »-l rH tH! lH5 «H) i-l r-» ,H ,H r-l rH |H «H «H; r^ r-t lH fH. iHiH
r-iiH v» CMCM enen mio i-ii-
-------
ITJ
4-1
03
a
M
•H
(0
a
$
rH
(0
•H
U
M
0)
0
u ;
•
HI
M
M «
(
M
,
1
<
1
i
<
f
•1
!
(
3j (
W «
'
C,
1
I
1
G
iN
• &
2«
8
CO
0 *
* I
i t
C
\
jj
*
a '
" S
*J
• Jj
1 Azrizoi
[Loaded
g
4
i*
1
j
3
<
i
M
(
1
<
c
1
1
(
I
§
t
J
3
8
f*
3
8
•
8
8
4
!
4
|
1
2
^
1
|
w
J
•
j
«
« «
v co
^ «
w •
CO i-l
o o
0 0
-------
ASM/IM240 Phoenix Lane Data Request Form
Please Fax or Mail Your Data Requests to:
Attn: William M. Pidgeon
U. S. Environmental Protection Agency
2565 Plymouth Road
Ann Arbor, MI 48105-2425
Phone #: (313) 668-4416
Fax #: (313) 668-4497
Requestor:
Organization:
Phone #:
Fax #:
Mailing Address:
4
Specify Disk Format
(3.5 inch only)
EH High Density
Specify File Format:
pi ,BM r-i M
U IBM U Mtc
Low Density
Text ASCII File
(tab separated)
CSV ASCII File
(comma separated)
SYLK File
Lotus File (.WKS)*
Excel File (v2.1)*
4
*Note: Oldest available formats were chosen for maximum compatibility.
These formats should be compatible with newer versions.
-------
Appendix C:
QC Steps for ASM Analysis Database
-------
The phoenix lane data used in these analyses were reported to EPA by the
testing contractor as total values (concentrations, mass, or flow) for the
entire test mode as well as in second-by-second form. The following automated
quality control (QC) checks were performed by EPA on the data. Tests that
were flagged by one (or more) of these QC checks were then manually verified.
Second by Second ASM Tolerance Check*:
• Speed Tolerance - ± 15% of nominal speed for Modes 1,2,3. Allowed
tolerance to be exceeded for less than 3 seconds in duration. Also
checked Idle for Modes 4,5
• Mode Length - Checked to ensure that each mode contained at least 20 and
not more than 30 "stable" seconds.
• Hp/Torque Tolerance - Compared required and actual horsepower (Hp) and
Torque and flagged differences > ±10% for at least 5 seconds.
All vehicles with test weights above 4000 pounds exceeded this tolerance
because of the capacity of the dynamometer. These cars were not removed
from these analyses. Smaller vehicles exceeding this tolerance were
removed.
• Calculated average concentrations and cumulative purge for all ASM
modes. Average concentrations were calculated as the average
concentration from second 10 to second 39 of each mode. The first 10
seconds of each mode were ignored to allow for the dynamometer
stabilization and exhaust transport time, vehicles with less than 30
seconds per mode were noted and vehicles with less than 20 seconds per
mode were excluded. Purge values were calculated as the total purge in
liters over the entire ASM including transient accelerations.
Second by Second ZM240 Tolerance Checks:
• Speed Tolerance - ± 4 mph at ± 1 sec of nominal speed. Allowed
tolerance to be exceeded for less than 3 seconds in duration. Also
speeds exceeding 70 mph, and less than 0 mph were flagged.
• Background Concentration Tolerances - Flagged background readings
outside the following ranges:
1.8 < HC < 10.0 (ppm)
-10.0 < CO < 30.0 (ppm)
- 0.5 < NOjc < 1.25 (ppm)
0.0 < C02 < 0.15 (percent)
• Test Length - Checked to ensure that the full 240 seconds were present.
• Distance Tolerance - Flagged distance > ± 5% of nominal distance
Bag 1: 0.532 < dist 1 < 0.588
Bag 2: 1.393 < dist 2 < 1.469
C-2
-------
• Fuel Economy Tolerance - Flagged fuel economies < 10 mpg and >50 mpg
• Sample Continuity and Integrity - Ensured that the sampling was
continuous (i.e., sec(I) - I for I - 1 to 240) and that gram and
concentration values were non-zero (HC, CO and C02 cannot all be zero
for fuel economy calculations or dilution factors).
Non-zero concentrations were not mandatory for the Phoenix data because
second by second concentrations received were calculated, not measured.
The calculated concentrations were based on the reported grains per
second results. These vehicles were still flagged for low
concentrations but were not removed from the analyses for this reason.
• Comparison of composite and bag results calculated from the second by
second data with composite and bags results received from ATL.
Differences of > 10% were flagged.
Purge Flow Data QC
• Comparison of second-by-second purge flow to the reported cumulative
purge flow and pass/fail status reported by ATL. All significant
differences were flagged.
• Vehicles exhibiting a non-zero constant purge rate for more than 20
seconds and at various speeds were flagged. Purge data was rounded to
nearest 0.01 liter/second prior to processing.
Bag FTP Tolerance Check*:
• The ratios of corresponding emissions (HC, CO, and NOX) and fuel economy
for each of the three bags that were not within expected ranges were
flagged.
• The temperatures, barometric pressures, and distances that were not
within expected ranges were flagged.
Bag IM240 Tolerance Checks:
• Bag-1 emissions (HC, CO, and NO*) and fuel economy were compared to the
corresponding Bag-2 results (based on regression analyses previously
performed on the Indiana data). All significant differences were
flagged.
• The Bag-1 and Bag-2 fuel economies were also compared to the test
weight. All fuel economy values that were not within an expected range
(based on test weight) were flagged.
C-3
-------
• The Bag-1 and Bag-2 distances not with the following ranges were
flagged:
Bag 1: 0.545 £ dist 1 £ 0.586
Bag 2: 1.365 S dist 2 Z 1.435
Bag IM240/FTP Tolerance Checks:
• For the laboratory recruited vehicles, the composite IM240 emissions
(HC, CO, and NOX) and fuel economy were compared to the corresponding
FTP results (based on regression analyses previously performed on the
Indiana data). All significant differences were flagged.
Dynamometer Loading Tolerance Checks:
• The test weights and horsepower settings had to be within 10% for all
tests performed on each vehicle.
Excluded Data Summary
This section of Appendix C details the vehicles excluded from the various
databases.
Purge: Analysis - 1725 of the 1758 lane tests contained the necessary
data to be included into this analysis. Of these 153 were removed because of
a malfunctioning purge meter, 184 tests were repeat tests for vehicles
previously tested and were removed, 5 cars had purge flow status fields which
indicated missing data, 95 additional vehicles had no indication of test order
and were removed, and 118 of the remaining vehicles exhibited non-zero
constant purge rates over varied vehicle speeds and were removed. The result
was a database of 1170 lane tested vehicles.
Outpoint Table Analysis - This analysis required laboratory FTP data.
Therefore, only lab recruited vehicles were considered for this analysis. Of
the 127 recruited vehicles, 17 did not receive initial ASM tests and one (veht
3258) did not receive an as-received FTP. Of the remaining 109 vehicles one
vehicle (vehf 2177) was removed because the ambient FTP temperature exceeded
allowable tolerances, one vehicle (veht 3253) was removed due to extremely
low HC emissions at the lane caused by a flame-out in the FID HC analyzer, and
vehf 3164 was removed due to unacceptable speed deviations on its initial ASM
test. The resulting database contained 106 lab recruited vehicles.
C-4
-------
Commercial Repair Analysis - Of the 27 vehicles recruited for this
program only 23 had completed after repair tests at the time of this analysis.
One vehicle, #13239 (CRt 24) was removed from the database due to unacceptable
speed deviations on its initial ASM test, leaving 22 vehicles available for
analysis. For the analysis of Section 5.6.2, 5 vehicles failed to pass the
Arizona state test on the subsequent retest and were removed. The resulting
database used for this analysis consisted of 17 vehicles which received
"successful" commercial repairs. For the commercial repair analysis of
Section 5.6.3, only vehicles initially failing ASM outpoints were included.
The result was 17 vehicles. Commercial repairs did not have to be successful
for this analysis and the two data sets contained slightly different cars.
Regression Coefficient Analysis - For this analysis all lab recruited
vehicles and commercial repair vehicles were removed from the analysis to
prevent the application of coefficients to data used to develop those
coefficients. Therefore, 1422 of the 1758 vehicles were considered for
inclusion into this analysis. Ten vehicles were removed because the composite
IM240 data was not available. The following vehicles were removed because
there was insufficient second by second data to calculate composite IM240
results:
Run t
1027
1855
2231
3066
3077
3079
3081
Reason for Removal
Test has only 93 seconds
Missing second by second
Test has only 93 seconds
Sampling Discontinuity
Sampling Discontinuity
Sampling Discontinuity
Sampling Discontinuity
Of the 1405 remaining vehicles, 1192 passed all QC tolerances. Purge
tolerances were not considered foe this analysis. The following table lists
the QC tolerances checks for which vehicles were removed from this analysis.
C-5
-------
Tolerance
riao/aed
ASM Speed
Short ASM Mode
ASM Horsepower
ZM240 Speed
IM240 Fuel Economy
IM240 Background
IM240 Sample
Nuaber of
Vehicles
8
2
10
14
4
163
18
Note: 1405 minus the above vehicles does not equal 1192 because some
vehicles exceeded more than one tolerance
Six hundred and eight (608) of the 1192 tests remaining received the IM240
second and were chosen for this analysis.
C-6
-------
Appendix D
IM240 Cutpoint Tablas
-------
5 s
ggooooooooooooooovovovooovomooooomomomotoio
voooooooooooooooooooooooodoooo'oodooo
oooooooooooooooooooooooooooooooooo
a
X
o
i
o
0)
in
ooooooooooo
ooooooooooooooooooooooooooooooooooo
oooooooooo
fl U
•3 ("• in m in in en en en en r* r* r^ r* fi fi
ainr>vor»r*voenenenooaBooeoaBaB
CbCMCMCMCMCMCMCMCMCMCMCMCMCMCMCM
a
9
H
2
£
•H
*
O
5
Q
M
1
!
5
S z
0
5 0
{! W
s
- X
_ Q
i z
'i 8
X U
I
9 U
^^•^•i^M
CM
i;
H
®
0
V
m
CO
CM
00
»
en
0
VO
x
in
0
in
CM
O
o
CM
m
to
ft
en
CM
CO
00
CM
CM
in
m
m
o
VO
g
O
CM
O
O
X
§
en
•w
CM
CM
00
CM
m
m
o
VO
X
CM
O
m
CM
0
o
o
o
m
in
ft
m
CM
CO
co
CM
CM
in
en
"»
•w
in
o
m
CM
O
oo
x.
o
m
m
ft
3
CM
CO
00
CM
CM
•n
en
e
CM
m
o
m
CM
X
O
m
o
CM
m
m
*•
CM
0
oo
aa
CM
VO
m
en
0
vo
in
o
CM
0
O
X
o
CM
00
m
m
en
s
en
CM
CM
in
n
0
VO
X
s
o
tn
x
0
o
o
m
VO
»
5
en
CO
CM
CM
in
in
m
vo
en
x
O
•f
CM
X
O
CM
X
s
in
VO
v
VO
an
oo
CM
CM
m
in
«
V0
A
•x
m
o
CM
X
O
CM
X
O
CM
in
VO
*
en
co
CM
CM
m
m
01
«*
CM
X
CM
VO
O
tf)
CM1
X
0
CM
X
s
in
vo
V
CM
CM
CM
in
m
en
H»
an
x
m
0
M
X
O
CM
X
O
M
m
00
CO
r»
CO
CM
CM
f
M
n
«
*
x
3
e
«
CM
X
O
00
X
s
in
00
3
r»
00
CM
CM
f
•n
n
0
CM
^»
O
«
CM
X
0
in
x
S
m
00
!3
oo
CM
CM
f*
lit
Ct
V
•w
x
O
K»
CM
•x
O
00
x
S
in
oo
3
OB
CM
CM
l»xxxxxx
OOOOOOOOOOOOOOOOO
CMCMCM>OOBaBCO»««voaBV*aBvOCOO
-------
v
OOOOOOOOOOOOOOOOOOOOOOOC40IC4MOOOOOOOO ~
a
X
o
ooooooooooooooooooooooooooooooooooo
• v0t0l0VOV0V0V0V0ak<00tV00«^9l>0O<'')£t0OOOO*0v0*0t0t0*a*0>0
yddodoooooooooooooddodddcNiMcNCNdddoodoo
c2
9
O
I
o
X
o^1^^1^1^^^
o a o o e
ooooooooooooooooo
o o o o e
ooooooooooooooooooooooooooo
eoinoo««>oooo«B«
^^'X^>»>.1^>,>»>i^>,
OOOOOOOOOOQO
(M d CM r4
OOOOOOOOOOOOOOOOOO
i
-------
i *
& ^
OOOOOOOOOOOOOOOOOOOf1r-r-<9coaoaoooo<.oao
i
^
fl (O t-l
5 in in
OOTO*OOOOOOF
co co co ^CNin(nmm«i»Jwo«»»wi~c->or— or— o«orr'itorfi»jr'«r«»»> in «i »i»i»>
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
-»--»--f> + -»'-f'-»'-f'-r'-»>* + -f'*-f'-»'-»'-*'>-»--t--«--f'-»--t>* + -*'*'f>'-»'**>
ooooooooooooooooooooooooooooooooooo
inoBincainoooo«>K>MO**iniAMr>«ooM-lOOOOOOOOOOOOOO
O
«0
o o o
«e va so
-------
•Mk
O
I
£ ° °
' 1 0 0
£
I » *
3 co co
a ° °
| V0 V0
£ vo vo
£ co co
M^H»
£ C- f
Zoo
CO OB
8SS
\£ |M
y oo oo
^H j^. .fc-
^^ *^l V'f
01 01
«s * *
Z CM CM
8 CO CO
CM CM
Ok Ok
in m
y o o
••BUB
O *
M m m
" CM CM
|«0
* tn m
V CM CM
I® °-
o oo
CM 0
o o
o o
I..
is=
0 0
0 0
CO CO
0 0
VO «0
co co
f r-
o o
CO 00
o o
00 CD
3 ot
« <0
CM CM
CM CM
Ok Ok
in in
0 0
O M>
8m
ot
m m
ot ot
o o
m ot
o o
o o
0 0
00 OB
O O O
0 0 O
.
o o o
OB CM CM
^O C^ C*l
^1 t^t ^1
00 CO 00
0 0 0
3 3 3
CM CM CM
CM CM CM
oo oo co
oo oo oo
CM CM CM
OB OB OB
m in in
£22
•N *• ^
So o
CD CD
O O O
000
0 0
0 0
V0 VO
o o
CM CM
CO CO
CM CM
CO CD
O O
CM CM
got
CD 00
s :
m in
*-4
22
SS
0 0
00 S
0
o
Ok
o
OB
co
oo
in
CM
CM
*"
Ok
M
-»
A
3
o
•*
o
o
Ok
o
CO
CO
00
m
CM
3
01
r*
•A
-
s
o
«t
o
"*
o
o
Ok
o
OB
CO
OB
m
CM
M
^
r»
K
8
o
*••
o
o
o
OB
CO
00
•n
CM
CM
***
Ok
M
5
s
0
*
o
0
o
CM
CO
00
in
CM
CM
Ok
Ok
r»
8
o
o
*
o
o
o
OB
CO
OB
•?
CM
CM
^
Ok
in
CO
o
o
~
••••
o
0
o
CM
OB
m
CM
CM
*^
Ok
in
s
o
0
o
OOOOOOOOCMCMCMCMCMOOOOOO
ooooooooooooooooooo
OOOOOOOOCMCMCMCMCMOOOOOO
««COCOCOCOrOCOOOBOBCMCMOkOkOkOICMCM
OBaoaocDOBCBaocoaBc»r»cococoa»oocooBa»
COM>M)OOOOCMCM
CMCMCMCMCMCMCMCMOOOOkOtmCOrOCOCMCM
CMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCM
*^* ^9 f4 eft l^ ^* ^* ^^ ^9 »^* o^* P§ ift •> lO lA Jrt *^ ^w
Vi •» (^ ^^ 1^ CO ^l ^^ In ^^ ^^ '•W »^* ^§ ^
-------
ii
I*
OOOOOCMCMCMCMCMCMeMCMOOOOOOOOOOQOOOC4CMCMCMr4CMeM NOi
3 28f
s Emissions
C CC
7 585
M S m
O
o
t-l
^^ ^4
J 0
1 ;•
1 :
s °.
3
o
m
o
1 i 2
0
O
00
00
CM
00
00
CM
CM
00
CM
00
m
0
o
oo
o
co
•v
o
CM
t-l
S
o
CM
0
o
00
00
CM
00
oo
CM
CM
00
CM
u>
00
in
r»
t-4
n
o
CM
•N.
O
CO
S
o
CM
O
o
00
00
CM
00
00
CM
CM
00
CM
Ul
00
m
0
2
In
o
CM
•x
0
m
o
\o
o
01
0
o
00
00
CM
00
00
CM
CM
00
CM
U)
00
m
p»
vo
<*»
CM
r-
O
en
•x
O
CM
S
0
A
i-l
O
CM
in
en
o
CM
00
o
r»
vo
t-i
00
CM
CM
P-
m
m
vo
.
o
o
t-l
o
VO
o
A
0
CM
m
*
CM
OO
CM
VO
CM
§
00
CM
OO
m
in
09
CM
VO
O
A
•-I
•x
O
o
o
o
t-l
A
t-l
0
CM
in
§
^
CM
00
CM
VO
»
0
O
o
CM
**
A
O
CM
m
A
*4
CM
O
VO
CM
O
*"
00
CM
m
vo
in
in
3
o
en
*v
o
o
t-l
o
o
ft
A
fl
O
CM
in
A
**
CO
o
vo
CM
§
co
CM
in
in
m
o
CM
•X
O
O
o
CM
t-l
A
t-l
oooooooooooooooooooooooo
CMCMOOOOOOOOOOOOOOCMCMCMcoooooaoaDoocoooa>aooor>r>a»aBooaoooao
CMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCM
•*f-t*4O.1>V-N.-X>«.^>.'X'X^>.>X*^*X'>«.>»'*«'^
oooooooooooooooooooooooo
oooooooooooooooooooooooo
v4*4OOOOOOOOOOOOOOOOOOOOOO
AAOOOOOOOOOOOOOOOOOOOOOO
r4v4
-------
acKcnenAen«oenenenenenenenen0«eneKcneKcnenenenenovotenenenencneftAen
^CMCMNCMCMCMCMCMC9OOOOOCMOOOOCMCMCMCMCMCMCMCMCM(MCMCMCMCMoic4
ooooooooooooooooooooooooooooooooooo
•0
c
•0
a.
X
a
• <«enenenehenir>ir4C\ii-4.-<<-li-lr4.-4VOr4l-lr-l,"4VOtaVO
X
II
u
01
a
CD CO CO CO CO
ft fl ft ft ft
VYOOOOOOCMOOOOOOOOOOOOOOOOCMCMCMCM
o«nfOcocofOco»«t>fo«nrof><*><»)<'i u> «e <-«
_. ckeMcnenencnoooocDaocoaDaBoaoooBa»i-ir>r>r>vovo<«vovovoiA>evor>r~««r«r>r-r-
CM CM CM CM CM CM CM
CMCMCMCMCMCMCMf^«n«n«»>C»»«nn«»»CMCMCMeMCMCMCMCMCMCMCMCMC>li4<-ioinr>r>r>i»r>i»c>)CMCMCMCMinininmininininincMr«c>><«m<«vo
CD CD OD CD oo oo oo ^ en en en en en en en o o o o OB CD oo do oo OB QD oo oo f> ft en ^4 ^4 ^ *+
inmmmmininiAinmK>ininininwv>u>Q)inininininininininv>«oin«e«a<««
rip^r^r^r^i^r^r^infl^^**1*
Si0vov0<0v0tawi0r»r»r-
i**' c^ co co co co co co c*o co co wO
>-«fir4f-««Dcocoa»r>i^(~r~r-r>r>i—r-r«CMi-,XiX>>.X.X,>,X,X,XJX,X,X1X,X»>»>»X.X^X.>.
OOOOOOOOOOOOOOOOOOOOOOO
incMOCMCMCMCM^»-toO'Hfioooenenoencnenen
ooooooooooooooooooooooooooooooooooo
I
CD A MY in
ddoooooo
ooooooooooooooooooooooooo
OOOOOOOOflftftftftfl«4ftftftf)^^^^4t^flflftflCMCMCMCOCOrOin
CMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMOtCICMeMCMCMCMCMCMCMCMCMCMCMCMOl
-------
*c CM Ol O O CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM (O (O CM CM CM CM CM CM CM CM CM CN
&N
s
£ooooooooooooooooooooooo25o°oooooooo
a
£
•n
OB
uCMCMOOCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMOOCMCMCMCMCMCMCMCMCMCM
c3
•o
c
•o
Q.
X
1)
u
O
*~*
in
c
0
l_(
o
0)
01
OOflOflOOOOOOOOOOOOOOOOOOOOaDOBOOOOOOOOOO
O
V0
I
Q
M
*»cocof'>c'»«ncofn(n<'>c'>cnr»r-f-r«r»r»t»r-
2 o o r» P»
at at ao ao
^»«n«n«nco«»»cn«nco«ncn«n(nfo
BfococofofOfotntncnrtMfnfofnf'tcMNCMCMCMCMCMCMCMCM'i'w^'w***^'*''!'
atatatatatatatatatatatatatatatatataiatatatatatatatatatatatatatatatateK
••^IB
!*ooo>atooooooooaor«i<>>r>r>r»r*«a»aiatatatatetaiatataioooooooo
Q*4,4Ototoo«aocoooooaocooeoo«oototototototO!O!Oiotoo*4r4'4>4<-i>4<-ir4
ZOOrOCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCNfMfMCMCMCMCMCntnfnfOfnOlinPltntO
••^•v«ri^inmminin«M«MCMC<««M(Makataiatataiaiati-ir4>«\Aatatatatatatatci>
inr>r*r>r>ri>inininininincMCMCMCMCMCMCM-«F4t-4atatatatakakatatatatcROQoooQoooowve«0>«vo««va«a«ovo«0<0«o«0<0«o<0vo
1
&
UCMCM^4«4r4^4>4i^«4^4»4v4p4^4v<00
*
00 00 00 0» 00 «0 ««
C^ f^ ^^ CO CO CO CO
o o o o o
oo oi oo ot CD
o o o o o o o
OD CM oi oo oo at at
o o o
oo o o
o o o o o
*0 r* *tf *o *£
o
I
a
OOO
0OOO
OOOO
C9OOOeOO
«9Oe90
^ ^4 ^ ^ ^ ^ ^ Cf W €*• Ci d Ol
O O O O O O
oooooeoo
ooooooooooooooooo
d^^docMCMcoaocftcDooaotao'oDco
i
ooooooooooooooooooooooooooooooooooo
'ddooodddoodoooooddddddddddododoodo
CMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCMCM(M(M(MCMCMCMCMCMCMC<4CMCMCMCM
-------
oooooooooo
o o o
o o o o o
o o o o o o
oooooooooo
•^•XV.'S.'X.^'X.'X.N.'V.
oooooooooo
Ul0tUI0kUI***
oooooooo
IS> O •- 10 O
OOft^MOfOH*
ooooooooooooooooooooooooooooooooooo
>^>»'S.>v.1S,'N'S,«V'>V>v'»»'>.*»«^^V'V'S..'XXi'X>.^^.1^^.^.'>.»-I— l-»l-«^l-l-'l-'»-'^t-'l->>-«(-»Mh*l— >-»l-»MI-'t->»-»l->ls>MM»O|iJ|>J
ooooooooooooooooooooooooooooooooooo
*
ooooooooooooooooooooooooooooooooooo
B
0k 0k
M
0
o o o o o o
2
CD » W CO
M
O
re
(5
n
ID
X
n
oooooooo
ooooooo«eoooooooooooooo
o o o o
10 a>
oooooooo
0
X
T3
B>
3
a
1
-------
o o o e e
eeeoooooooeooeooooooooooooooo
UlbJU>WOIU>UIUIU>U)U)U)U>WWU>b>U>UICt>UIUIUI
ooooooooooooooooooooooo
09 10 flB « OB <0
UI4b*>fa>fa»WU>b>U>U>W
ooooooooooo
xx,xx,xxxxxxx
O»O>tOO»»IOO»-'t<>OM
ooooooooooooooooooooooeeeoooooeoeoo
-^XXXXXXXXXX-^XXXXXXXX>X.XXXXXXXXXXXX.XX
^^,^l_>,_>»«»->t*»<»*l_>l_>l->h*»*l->»->l->**»jMlotOI->l-*l-'l->l->l-«H>l->l->»->IOIO
u>uiuiuiuioio*«J^0tvioi«J«>i«JCDaBtowoo»*»>>uiuiuiuioBOBOBtowJO»woo»too0k»J0kw0t«j0k«j
ooooooooooooooooooooooo
o o o
O OB JO O OB SO
o o o o o o
MIO«d«J«J
24ft 4ft 4ft ^^ 1^ (ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft
»»***UimUIUtUIUlUIUI«»»»*MMIOIOIOIO»»>IO
U>»»»«MI-*l»H<»»OOOOOO**»*»»*»W«Ju)u>u>uiu>wi<)«j->j^»>tOIOIS»IOIOMMMIOIOIOIOOOOOOOOO
OOOOOtOMMMMMIOIO^'k'MI-'OOOOOOOO
tototototototototototoiotototototo«oioio>ioioioioioio
»**J«J>JMIOIOM0teO«IOIO»>>IOMIOMM«>**»UIUIUIUIUIUIUIUI
tototo«ototetotetotote«otetotototoiototoioiowtetototooaaBaBo»soBaBO9 =
- ~ ~ ~ ~ ~ ~ •k*«»*>»*UtUIU>W4>J->o"io>oi*->->-'»-'e'>oNO>ioie>>o>ooo
OOOOBOBO»OBUIUIUIUIUIUIUIUI
E
*J ^J (ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft Ift 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft 4ft
uju>H'»~H'»~H<»<**io»tototo
StOtOtOtOtOMtOtOtOVDtoB.
aBo»o»aBaBO>eDOBO»OBO>p
o
V.
10
s>
n
o
3
x
•o
a\f>0>O\O\0\O\0>O*O\0i9iO\Oi9\0iO>9*9>9\O>99>0>&O\9i0>&>0*0>O*&'O*
ooooooooooooooooooooooooooooooooooo
fOMIOMIOIOIOIOIOIOIOIOtOIOIOMMIOIOIOIOIOIOIOIOMMIOMIOMIOIOIOlO
tetotototototototototo«o»«w«to«toto>etototototoIOfO**Jb*IS>|OlOlS>fOK9lOK>
V>^OtO^Oti9tO^OtO^O>OtO^O^O^O^O
M
-------
o o o e o
o o o o o
oppppppppppooooopopooooooo
oeoooooeoooooooooeoooooooo
ooooooooooeoooooooooooooooooooooooo
V.X.X.X.X.X.X.X.X.X.X.'X.X.V.V.'N.'VV.X.'VX.V.X.X.X.XV.V'X.X.X'X.X.'V'V.
MMMMIOMMMIOM
•W WW W^ w r^ ^M _^A
M ^ U ^ ^ f f
uiuimwiuto\o»et0\0i->J-J-J«J«j«aBOD»aDwwww>woooo
•*- + <4- + + H-'f + 'f* + + 't-* + * + -«. + ******'«-* + * + ** + * + 3J§
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOS
ooooooooooooooooo
o
I
g s s s
$ s s s s s g s s s s s s s $ s s s
oooooooooooooooooo
...
000"
Mi^MMMooooooooooooooooooooooo
b b b b O
X
(t
r
I
o
n
ooooooooooooooooooooooooooooooooooo
(8
X
•o
OD OD
oooooooooooooooooooooooooooooo
i
a
n
.
ooou>wooooooooooooooooooooobbbbbbboo»g.
-------
•»iq»I
-------
HC FTP (g/mi)
o
o
E
o
o
« r*
^ y
P- °
O
o
to
o
B
O
BC FTP (g/mi)
o
o
N>
O
o
o
o
—I-
to
o
o
•H
ro
o
o
o
CO
o
-------
CO FTP
i-»
o m
K)
O
s
8
N>
o
K>
-H
8
CO FTP (g/ai)
o m o
to
Ul •
8
to
o
-I 1
8
CO
I
B
O
ft
-------
MOx FTP (ff/«i)
I
e
1
es> • •
t u, -.
10
•
tn
•H
8
o
MOx rPP (g/ml)
o
u« M
t I
NJ
in
1
10
CO
S
Sfl
1
10
-------
Appendix G:
ARCO, Si«rra, Environment Canada Data Analysis
-------
i . o
The objective of thia report is to respond to the pilot ASM teat
programs performed by Sierra Research, Inc., ARCO Products Company, and
Environment Canada. Sierra and ARCO both previously published papers
praising the capabilities of the ASM, and both concluded that some form of the
ASM could replace the IM240 as an enhanced I/M test.
EPA has concluded that the ARCO and Sierra reports are incorrect in
claiming the ASM as equal to the IM240. Based on a comparison with a similar
database of IM240 vehicles, the ASM is inferior to the IM240 at identifying
excess emissions without committing false failures. Moreover, a series of
regressions were run for both the ASM and the IM240 versus the FTP. The
scatterplots for these regressions, contained in the Appendix to this report,
show significant variability for the ASM at predicting FTP values, compared to
the IM240.
A contractor for EPA is currently testing a number of vehicles at a
state I/M lane in Mesa, Arizona on both the IM240, and a 4 mode steady state
test, which includes two ASMs, the ASM2525 and the ASM5015. A sample of
vehicles is being recruited to the contractor's lab for further FTP testing.
The data from that program will give EPA a chance to determine, with greater
confidence, if some form of the ASM is as effective as the IM240.
This report focuses on a small dataaet of vehicles, therefore the
conclusions made in this report are subject to change when more data is
available to EPA. However, from the data that has been presented to EPA to
date on the ASMS, the ZM240 remains the only enhanced I/M test.
2.0 Database Peaertpfcian
There are 31 vehicles in the ASM database EPA used for this analysis.
The data were gathered from programs performed by three different
organizations: Environment Canada^, Sierra Research^, and ARCO Product s^.
*• Ballantyne, Vera Jf. Pr«<». St-^ady St-at-» Tgatirny tUtparfc and Dafca.
Environment Canada, August 28, 1992.
2 Austin, Thomas C., Sherwood, Larry, P»vi»lnpm*»nr at Tmprovad Loaded—Moda
Teafc Praexiufaji £oy Ingpaeti.Qt^ and Mainfftnanra Prnrrrama. Sierra Research,
Inc. and California Bureau of Automotive Repair, SAE Paper No. 891120,
Government/Industry Meeting and Exposition, May 2-4, 1989.
G-2
-------
EPA started performing ASM teats in Mesa Arizona on September 10, 1992. These
data will be the topic of a separate analysis.
A number of vehicles in the ASM database were tested with and without
implanted defects, so 51 test configurations were used for this analysis. All
the vehicles tested by the three different organizations received the ASMS015
and the FTP, but ARCO did not perform the ASM2525. This left 39 test
configurations receiving multiple-mode ASM tests and FTPs.
2.1 ASM Vehicle* Removed from Database
There were originally 55 vehicles tested in the three programs, resulting
in 117 teat configurations, broken down as follows: Environment Canada (32
vehicles or 36 configurations); Sierra Research (18 vehicles or 51
configurations), and ARCO Products (5 vehicles or 30 configurations).
Vehicles were removed from the database for reasons which are discussed below.
First, all pre-1983 vehicles were removed to focus on newer
technology vehicles. So 3 Canadian vehicles and 5 Sierra vehicles were
removed, leaving 29 Canadian vehicles with 33 configurations and 13 Sierra
vehicles alao with 33 configurations.
Next all pre-1988 Canadian vehicles were removed. Canadian vehicle
standards were not lowered to 0.41/3.4/1.0 until the 1988 model year, so the
prior model years could not be used. So 13 Canadian vehicles were removed,
leaving 16 Canadian vehicles with 20 configurations.
Next, all ARCO vehicles that were not certified to the 50-state standards
of 0.41/3.4/1.0 were removed. Three ARCO vehicles were certified to
California-only standards, so they were removed, leaving 2 ARCO vehicles with
12 configurations.
Finally, all Sierra configurations that received hot-start FTPS instead
of cold-start FTPs were removed. Because the normal cold-start FT? is more
variable than hot-start FTPs, short test comparisons should be mad* using
cold-start FTPs. Also, vehicles are certified using cold-start FTPs, so the
results are more relevant. So 14 Sierra configurations were removed, leaving
13 Sierra vehicles with 19 configurations.
BoelchauS Kenneth L., et al. Evaluation at Enhanearf Tn.ip^efclon Taehnlaruaa on
Automobiles. ARCO Products Company Report, May 8,1992.
G-3
-------
2.2
Selection of IM240 vehicles Used ia Database
In order to compare the ASM to the IM240, the analysis should be performed
on a set of vehicles that have received both tests. However, none of the ASM
vehicles received the IM240, therefore 39 vehicles were randomly selected
from the Indiana laboratory IM240 database. These vehicles were chosen from
those used in the IM240 cutpoint table analysis in EPA's I/M Coa<-«. Benafita.
and Impacta Analyaiar which included 274 vehicles with both IM240 and FTP
results. In order to make the IM240 database similar to the ASM database, the
following process was used.
First, the ASM vehicles were categorized by emission levels according to
the following table:
Table 1
Miunbar of Vahlelaa In
oar Emittant
HC/CO
Category
Normal
Normal
High
Very High
Very High
NOx
Category
Normal
High
Normal
Normal
High
HC Range*
0£HC<0 . 82
0£HC<0 . 82
0.8252KX1.64
1.64£HC<10.0
1.64SHC<10.0
CO Range*
0£CO<10.2
OSCCK10.2
10.2SOX13.6
13.6SOX150
13.65CO<150
NOx Range
0£NOx<2
253IOx<4
O^NOx<2
OSNOx<2
2SNOx<4
f ia
Dataset
29
1
2
4
3
* These are the same categories as those used in the I/M Technical Support
Document
Second, the Lab IM240 database was broken down into these same categories.
All vehicles were 1983* model years, and only vehicles that received the lab
IM240 after the FTP were kept in the database. This kept the IM240 database
as similar as possible to the ASM database. From the remaining vehicles, a
random sample waa chosen from each category so that both databases had the
same number of vehicles in each category.
By selecting the same number of vehicles from each emittant range, it
prevents one test from getting an unfair advantage in achieving identification
rates. For example, if the ZM240 database included considerably higher FTP
scores, it would have identified much more excess emissions, thus making its
Identification Rates (IDRs) higher.
G-4
-------
3 . 0 g»lgul«fclner ASM Maaa
Sierra indicated (SAE Paper No. 891120) that calculated ASM mass emissions
correlate better to the FTP than concentration measurements, so their method
of converting ASM NOx concentration measurements to "mass" emissions was
applied to this ASM database for HC and CO, as well as NOx. This was done by
multiplying the emission concentrations (ppm for HC and NOx, and % for CO) by
the vehicles' Inertia Weights (IW), yielding the following units: kiloton-ppm
for HC (IW * ppm/103), ton-% for CO (IW * %), and megaton-ppm for NOx (IW *
ppm/106) . These are the values EPA used for the regressions in this report.
3.1 BPA Equation* Versus Sierra equations
In their test program, Sierra measured the ASM emissions on both a
concentration basis and mass basis. This allowed them to regress
Concentration * Inertia Weight (IW) versus mass emissions for the same test,
and develop equations that convert (Concentration * IW] to Mass. As expected,
these mass calculations correlated very well with the measured mass emissions.
Sierra's next step wss to regress the measured steady state mass emissions
against the FTP emissions and report r^s for these regressions. They did not
actually use the calculated mass emissions to predict FTP scores. This is
where EPA's analysis of the ASMS was slightly different. EPA regressed the
[Concentration * IW] values against the FTP emissions for each vehicle. This
was done because EPA did not have measured mass emissions from all three test
programs compiled in this report. However, the major benefit of the ASMS,
according to Sierra and ARCO, is the ability to use the less expensive BAR90
type analyzers when measuring the exhaust concentrations. Sinca this is a
claimed benefit of the ASMS, the readings from these less expensive analyzers
should be used when comparing the ASM to the IM240.
4 . 0 Mnltelpla t.ln^»g ft*ag«««tona tor th*
Using data from all thre* previously mentioned programs, EPA calculated
the IW * Concentration for each emittant. Then a multiple linear regression
was performed, using th« calculated (Hf*Concentration) ASM2525 and ASM5015
scores as two separate variables vs measured FTP emissions. Equations were
developed from these regressions that predict an FTP score from a combination
of the ASM2S25 and ASM5015 concentrations * IW scores:
G-5
-------
Table 2,
Bemafc ieina Developed r.o Pr»rffr!«- FTP from ASM Modes.
Predicted FTP - [IW (A*ASM2525 -t- B*ASM5015) + C]
Bmittant
HC (ppm)
CO (%>
NOx (ppm)
A
-3. 96xlO-7
2.64xlO-3
1.13x10-7
B
4.60xlO-7
S.lOxlO-5
1.30x10-7
C
0.523520
4.222840
0.515531
r2
49.2%
43.5%
71.4%
4.1
Simple Linear Regressions
Aside from those already mentioned, regressions were also run for each
individual ASM mode vs the FTP, and for the IM240 vs the FTP. Since the IM240
is a transient test, like the FTP, it correlates much better to the FTP than
the ASM modes.
4.1.1 Coefficient of Determination (r2)
The r2 may be interpreted as the proportion of the total FTP variability
that was predicted by the short test. For example, if the r2 equalled 100%,
the short test would have perfectly predicted the FTP scores for these cars.
If the r2 for these vehicles was zero, the short test would not have any
linear relationship to the FTP.
The r2 data, listed in table below, show that the IM240 is considerably
better than the ASM tests in predicting FTP HC, CO, and NOx scores. For HC
and CO, less than half of the FTP variation is explained by the ASM scores.
Table 3.
St-ar.iar.ieal C.emtetMfiaon of r.h» FTP Versus I/M Tests
r2
EC
IM240
95%
5015
36%
2525
20%
CO
IM240
92%
5015
45%
2525
44%
NOx
IM240
84%
5015
62%
2525
70%
4.2
Scatterplots
For an I/M test, more important than the r2 ia the ability to identify
high proportions of dirty cars without falsely failing vehicles. The IK240
also has a significant advantage at identifying more of the dirty cars while
failing less of the clean cars. The scatterplots in the appendix show this
clearly. When viewing the plots, consider the following chart for a
reference.
6-6
-------
Used to Pefina Terma
5 y
4 . .
5 Excess
-*. * Emissions
~ Omitted ^»
E 2" ^^^
Clean Cars
.sip&aa both tests)
^
^/^
^)X^xcess
^^^ Emissions
*r Identified
Errors of Commission
01234
Short Test Score
The short test outpoint under consideration is the vertical line, and
the FTP standard is the horizontal line. The intersection of these lines
splits the chart into quadrants. The goal of the short test is to maximize
the number of FTP failing vehicles into the upper right quadrant, while
minimizing the false failures in the lower right quadrant.
The more vehicles that appear in the upper left quadrant, the less
effective the test becomes, because these are all dirty cars that are not
identified by the short teat. From this perspective, the advantages of the
IM240 is clear. Every IM240 chart shows that an x-axis value (outpoint) can
be selected that clearly places the vast majority of dirty cars in the upper-
right quadrant, without errora of commission. The ASM tests do not display
this trait nearly as well. Only the 2-mode ASM tests and the ZM240
scatterplots have the horizontal and vertical lines on them, so the reader can
examine different outpoint scenarios.
G-7
-------
4.2.1 Scattarplot Statistics
Each of the regression scatterplots contains the following information:
• The best-fit regression line showing predicted FTP for a continuum of
short test scores, developed from a regression of the actual data.
• 'Boundary Lines' at + 2 and - 2 standard error from the predicted
value.
• A horizontal dotted line at the FTP standard.
• A box containing descriptive statistics.
On each Regression plot, a box in the upper-left corner provides the
following statistics: 1) The equation of the line used to predict FTP values
from the short test's score. 2) r2, discussed above. 3) The standard error*
, a statistic that describes the variability of the FTP score predicted from
the selected short test. The next section discusses standard error in more
detail.
4.3 Standard Krxox aa a Measure of Variability
The weakness of the ASM testa regarding r^ and the low proportion of
cars that can be identified aa dirty while simultaneously avoiding false
failures, is related to teat variability. The standard error is an objective
measurement of test variability. The following shows that the ASM tests are
significantly more variable than the XM240, using the standard error as an
objective measure of variability.
4.3.1 Aaauaptiona Mad* for Using Standard Error
The following assumptions were made in order to use standard error as it
is used in this report:
• Linear relationship between the FTP and the short testa.
• Normally distributed data.
• Homoscedaatic diatribution (i.e., the standard deviation of FTP values
is constant for all abort teat values).
What is referred to in this report is formally termed standard error of
estimate, but for convenience purposes, will simply be called standard
error.
G-8
-------
The standard error is similar to standard deviation because a bandwidth
of ±1 std. error includes -68% of the data and ±2 std. error includes -95% of
the data.
4.3.2 Example Using Standard Error
Consider a 3000 Ib. vehicle that emits 1500 ppm NOx on both the ASM5015
and ASM2525. Plugging these numbers into the equation for predicting FTP
values (Table 2) yields 1.61 g/mi. However, because the standard error for
ASM NOx (see Table 4) is 0.36 g/mile, roughly 5% of the FTP scores predicted
by the ASM result will be greater than 2.33 g/mile (1.61 + 2*0.36) or less
than 0.89 g/mile (1.61 - 2*0.36). Since half of these will err on the low
side, it is probable that -2.5% of the vehicles identified as failures by an
ASM cutpoint of 1.61 g/mi would be false failures.
4.3.3 Effect of Standard Error on "Safe FTP Predictions"
In order to be confident the false failure rate would be less than 2.5% the
selected cutpoint should predict an FTP value of 2 standard errors greater than
the FTP standard. This ensures that the low values (FTP prediction - 2 std.
error) are still failing the FTP.
For example,
FTP NOx standard is 1.0 g/mi
The ASM NOx std. error is 0.36 g/mi
FTPstandard + 2 std. errors -1.72 g/mi
So, the selected ASM cutpoint should predict an FTP of no less than 1.72
g/mi. Applying the same logic to the IM240, whose standard error is 0.28 g/mi, a
predicted FTP score of 1.56 g/mi (1.0 + 2*0.28) will also yield an error of
commission rate leaa than 2.5%. But because the "safe* predicted FTP score is
more stringent, the excess emissions identified will be higher. The standard
errors and predicted FTP levels that are expected to limit false failures to
approximately 2.5% are compared in Table 4 below.
G-9
-------
Table 4. Comparison of ASM and TM24Q St-anHard erraca And Their Effect
Prodletad FTP Sfrrlnaanev at a 2 51 Falsa Failure
1 std. error (g/mi)
Predicted FTP Level
8 -2.5% EC (g/ mi)
HC
IM240
0
0
.24
.89
ASM
0
1
.60
.61
XM2
3.
11
CO
40
8
.0
ASM
4.
13
8
.0
NO*
IM240
0
1
.28
.56
ASM
0.36
1.72
5 . 0 gutpatnf
Another way to assess the effectiveness of I/M tests is to evaluate the
following factors, which were discussed in detail in Section 4.2.1 of EPA*s
T/M Coats. Bftnafltsr and Imparts Analysts; excess emission identification
rates, failure rates, error-of-commission rate, the failure rate among
vehicles that pass FTP standards, and the failure rate for so-called "normal
emitters," which may fail an FTP standard (normal emitters are defined as
vehicles whoa* FTP HC < 0.82 g/mi and FTP CO < 10.2 g/mi), but are clean
enough to make the coat effectiveness of repairs an issue. These factors are
highly interactive, for example, high IDRa can be achieved with stringent
outpoints, but this will adversely affect failure rates.
Cutpoint tables for the ASM tests and the IM240 in the appendix allow
these factors to be compared. The cutpoints for the tables were chosen using
an iterative process. The goal was to select cutpoints that would give
reasonable identification rates while limiting errors of commission. The goal
was to keep the EC rate at 0% for both procedures.
For both cutpoint tables, four different cutpoints were selected for each
of the three emittanta, resulting in 64 different cutpoint combinations. For
the IM240, the "Two Nays to Pass Criteria" was used, as described in Section
4.2.3.2 Of EPA* 8 T/M Coats. Banafits. and Impacts Analysis. Thia ia a method
of combining the compoaite HC and CO acorea with the bag 2 HC and CO scores in
order to minimize Errors of Commission on vehicles with cold start problems,
while maintaining high Identification Rates.
5 . 1 Selecting ASM Outpoint*
Scatterplota were done plotting Measured FTP vs. Calculated FTP from the
ASM scores. From these scatterplots, EPA determined a range of cutpoints to
use for the cutpoint tables. For example, looking at Chart x, FTP CO vs ASM
G-10
-------
Prediction CO, it can be determined that an ASM Prediction between 6 and 10
grams/mile would identify most dirty vehicles (those above the standard of 3.4
g/mi) without failing the clean vehicles. It is also obvious that a outpoint
of 5 would falsely fail at least one vehicle while achieving no added benefit.
Consequently, the chosen CO outpoints for the ASM range from 6 to 20. The
range of outpoints for HC and NOx were chosen the same way.
5.1.1 Using Standard error to Predict Reasonable Outpoints
Although the cutpoint tables do include a wide range of outpoints, there is
still a concern that the errors of commission are not representative of what they
might be in a real world scenario. For this reason, the outpoints shaded at the
end of each table were selected using the standard error.
The ASM cutpoints used are identical to the "safe FTP predictions" in
Table 4. This is because the values are obtained from calculations using both
ASM scores. Each mode has a "sliding scale" of cutpoints, dependent on the other
mode results. In other words, no single ASM5015 or ASM2525 value can be used for
a cutpoint since a vehicle might be clean on one mode and very dirty on the
other. The cutpoints for the IM240, on the other hand, are direct IM240 scores,
in grams per mile. The "safe cutpoints* for the IM240 were determined by
calculating the IM240 score that would predict the "safe FTP Level - 2.5% EC"
(see Table 4), using the equations on each respective IM240 scatterplot.
For example, the IM240 HC FTP Level - 2.5% EC is 0.89 g/mi. The regression
equation on the IM240 vs FTP scatterplot is:
- 1.429 * IM240 + 0.04;
Since we want to predict an FTP of no less than 0.89 g/mi, setting FTPpced.
equal to 0.89 yields an IM240 score of 0.60 g/mi. This was done to calculate
each "safe cutpoint" for the ZM240.
5.2 Limitations of Cutpoint Table*
It ia important to recognize several limitations in these tables. Most
important is that the database ia very small and does not represent the in use
fleet. Additionally, the vehicles were preconditioned by the FTP before the
ASM test and before th« IM240 testa, so the correlation between these short
tests will be much better than can be expected for vehicles tested in an I/M
lane, because of all the uncontrolled variables associated with I/M lane tests
like temperature, fuel RVP, distance driven prior to the test, catalyst
temperature, etc. Because all of these variables were controlled for the
G-ll
-------
vehicles in the ASM and IM240 databases, the cutpoints can be very stringent
while still avoiding false failures. For example, the ZM240 table shows that
cutpoints of 0.4/6/1.0 yield IDRs of 97%, 93%, and 100%, for HC, CO and NOx,
respectively, without errors of commission. If cutpoints this stringent were
used for random vehicles tested in I/M lanes, the error of commission rate
would be unacceptably high. Similarly, because of the introduced
malfunctions, the failure rates are not representative of the in-use fleet
failure rates for an acceptable I/M program. So, while it is valid to use
these cutpoint tables to compare the ASM to the IM240, it is not valid to
assume that the rates are representative of those that will be realized in a
real I/M program. The ASM and IM240 testing that EPA is currently sponsoring
in Mesa will provide the actual in-use fleet rates.
5.3 XM240 Identifies Much More Excess Emissions
Using the cutpoint tables to compare the two procedures, the IM240 did
considerably better than the Two-Mode ASM at each tests' optimal cutpoints* .
The IM240 identified 97% of excess HC, 931 of excess CO, and 100% of excess
NOx at cutpoints of 0.4/6/1.0 (HC/CO/NOx). The Two-Mode ASM identified 87%,
80%, and 75% of HC, CO, and NOx, respectively at cutpoints of 0.6/6/1.50.
As discussed in the Variability section, using the standard error of
estimate to choose cutpoints that should prevent exceeding an error of
commission rate of 2.5% can help in assessing the performance of I/M tests.
The shaded cutpoints at the end of each test's cutpoint table suggest that the
IM240's performance is significantly better than the two-mode ASMs. Using the
"safe" cutpoints, the IM240 identifies 92%, 84%, and 71% excess of the excess
HC, CO, and NOx, respectively - the Two-Mode ASM only identifies 75%, 63%, and
64%.
£.0
The ASM teats were considerably more variable than the IM240 under
controlled laboratory conditions, aa evidenced by subjective analyses of the
scatter plots and objective measurements using the standard error statistic.
Testing at real-world I/M lanes will add considerably more variability to both
tests, because conditions known to affect emissions such aa temperature,
humidity, and vehicle operating conditions prior to the test. These
uncontrolled variables are expected to add proportionally more variability to
* 'Optimal Cutpoints', aa used here, is the lowest cutpoints the test could go
to and still have zero errors of commission.
G-12
-------
a steady state teat like the ASM, but data are not available to evaluate the
validity of the hypothesis.
On the other hand, the increased variability associated with actual I/M
testing will be somewhat offset for the ASM by adding two additional modes; a
50 mph steady mode at road-load horsepower, and an idle mode. This four-mode
ASM procedure is now being performed by EPA in a Mesa Arizona I/M lane.
The result of these offsetting effects on variability will determine the
viability of the ASM as a lower cost substitute for the IM240. A final
conclusion should be postponed until enough Mesa data can gathered for a valid
evaluation.
G-13
-------
Appendix A
Outpoint Tables
-------
to VMM"
XMMO*
39 19*3+ ?"r*
Inalir tmlmr
IM240 Cvt-Poiot*
CMS) t 1M 2
0.4/15/2.00 +
0.8/15/2.00 +
1.0/15/2.00 +•
1.2/15/2.00 +
0.4/10/2.00 +
0.8/10/2.00 +
1.0/10/2.00 *
1.2/10/2.00 +
0.4/ 6/2.00 *
0.8/ 6/2.00 *
l.O/ 6/2.00 +
1.2/ 6/2.00 «•
0.4/ 5/2.00 *
0.8/ 5/2.00 +
l.O/ 5/2.00 +
1.2/ 5/2.00 +
0.4/15/1.50 +
0.8/15/1.50 +
1.0/15/1.50 +
1.2/15/1.50 *
0.4/10/1.50 +
0.8/10/1.50 *
1.0/10/1. SO +
1.2/10/1.50 +
0.4/ 6/1.50 *
0.8/ 6/1.50 +
l.O/ 6/1.50 +
1.2/ 6/1.50 *
0.4/ 3/1.50 *•
0.8/ 5/1.50 *
l.O/ 5/1.50 *
1.2/ S/1.50 +
0.4/15/1. 23 +
0.8/15/1.23 *.
1.0/13/1.23 *
1.2/15/1.25 +
0.3/12
0.5/12
0.6/12
0.8/12
0.3/ 8
O.S/ 8
0.6/8
0.8/ 8
0.3/4.5
0.5/4.5
0.6/4.5
0.8/4.3
0.3/ 4
O.S/ 4
0.6/ 4
0.8/ 4
0.3/12
0.5/12
0.6/12
0.8/12
0.3/ 8
O.S/ 8
0.6/ 8
O.S/ S
0.3/4.S
0.3/4.3
0.6/4.3
0.8/4.3
0.3/ 4
O.S/ 4
0.6/ 4
0.8/ 4
0.3/12
0.3/12
0.6/12
0.8/12
itexi Vm
^4 •
•
XdMt4flo*t*oa
R*t*
K
96%
91%
91%
91%
96%
91%
91%
91%
97%
94%
94%
94%
97%
94%
94%
94%
96%
91%
91%
91%
96%
91%
91%
91%
97%
94%
94%
94%
97%
94%
94%
94%
96%
91%
91%
91%
CQ
86%
80%
80%
80%
86%
82%
82%
82%
91%
89%
89%
89%
94%
93%
92%
»a%
86%
80%
80%
80%
86%
82%
82%
82%
91%
69%
8»%
89%
M*
M%
m
tat
87%
81%
81%
81%
ate
93%
71%
71%
71%
93%
71%
71%
71%
97%
75%
75%
73%
97%
73%
73%
75%
93%
71%
71%
71%
93%
71%
71%
71%
97%
73%
73%
73%
97%
73%
73%
73%
93%
73%
73%
73%
of
Hilan,
22
17
16
16
22
18
17
17
26
23
22
22
30
27
26
26
22
17
16
16
22
18
17
17
26
23
22
22
30
27
26
26
23
IS
17
17
B*a
48%
37%
35%
35%
48%
39%
37%
37%
57%
50%
48%
48%
65%
59%
57%
57%
48%
37%
35%
35%
48%
39%
37%
37%
37%
50%
48%
48%
65%
59%
57%
57%
50%
39%
37%
37%
of
Xft^
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
0
0
0
0
Bftfc*
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
2%
2%
2%
2%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
2%
2%
2%
2%
0%
0%
0%
0%
tailor* 8*to
foe R»
XtUolu
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
10%
10%
10%
10%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
10%
10%
10%
10%
0%
0%
0%
0%
fa&lvr* R*to
for NotMl
teittiag
UfcifllM
16%
3%
3%
3%
16%
6%
6%
6%
29%
23%
23%
23%
42%
33%
35%
33%
16%
3%
3%
3%
16%
«%
«%
6%
29%
23%
23%
23%
42%
33%
33%
33%
19%
6%
6*
«»
A-l
-------
"Tw« Itay* to *•••"
40*
39 1983* •— *
«lv a*lma
11840 Cw*-P«iat»
r.tma t IM a
0.4/10/1.25 *
0.8/10/1.25 *
1.0/10/1.25 +
1.2/10/1.25 +
0.4/ 6/1.25 +
0.8/ 6/1.25 *•
l.O/ 6/1.25 +
1.2/ 6/1.25 +
0.4/ 5/1.25 +
0.8/ 5/1.25 *
l.O/ 5/1.25 +
1.2/ 5/1.25 +
0.4/15/1.00 *
0.8/15/1.00 *
1.0/15/1.00 *
1.2/13/1.00 *
0.4/10/1.00 *
0.8/10/1.00 +
1.0/10/1.00 *
1.2/10/1.00 *
0.4/ 6/1.00 +
0.8/ 6/1.00 +
l.O/ 6/1.00 *
1.2/ 6/1.00 +
0.4/ 5/1.00 *
0.8/ 5/1.00 *
l.O/ 5/1.00 *
1.2/ 5/1.00 +
0.6/10/1.60 +
0.3/8
0.5/ 8
0.6/8
0.8/ 8
0.3/4.3
0.5/4.5
0.6/4.5
0.8/4.5
0.3/4
0.5/ 4
0.6/4
0.8/ 4
0.3/12
0.5/12
0.6/12
0.8/12
0.3/ 8
O.S/ 8
0.6/ 8
0.8/ 8
0.3/4.9
0.9/4.9
0.6/4.9
0.8/4.9
0.3/ 4
0.9/ 4
0.6/ 4
0.8/ 4
O.S/ 9
**4 v«
k4««l A*
Xdaatlfiiu
ac
96%
91%
91%
91%
97%
94%
94%
94%
97%
94%
94%
94%
97%
93%
95%
93%
97%
93%
93%
95%
97%
93%
93%
99%
97%
99%
99%
99%
92%
CQ
87%
83%
83%
83%
91%
89%
89%
89%
94%
92%
92%
92%
91%
90%
90%
90%
91%
90%
90%
90%
93%
92%
92%
92%
94%
94%
94%
94%
84%
•
itlea
BOB
93%
73%
73%
73%
97%
75%
75%
75%
97%
75%
75%
75%
100%
93%
93%
93%
100%
93%
93%
93%
100%
93%
93%
93%
100%
93%
93%
93%
71%
ot
laUasM
23
19
18
18
26
23
22
22
30
27
26
26
27
23
24
24
27
23
24
24
29
27
26
26
32
30
29
29
19
!•£•
50%
41%
39%
39%
57%
50%
48%
48%
65%
59%
57%
57%
59%
54%
52%
52%
59%
54%
52%
52%
«3%
59%
57%
57%
70%
65%
63%
63%
41%
e«
IflU*
0
0
0
0
0
0
0
0
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
0
1
••
Bate
0%
0%
0%
0%
0%
0%
0%
0%
2%
2%
2%
2%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
2%
2%
2%
2%
0%
ration ««t»
f o* ra
PflLAdkBV
Vftki^A^AiV
0%
0%
0%
.0%
0%
0%
0%
0%
10%
10%
10%
10%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
10%
10%
10%
10%
0%
f 0* M*CMl
tatttla*
S&biaiM
19%
10%
10%
10%
29%
23%
23%
23%
42%
35%
35%
35%
32%
29%
29%
29%
32%
29%
29%
29%
39%
35%
35%
35%
48%
43%
43%
45%
10%
the* ••!*«••*
thttt
th* •**•*(•) •« O
A-2
-------
AMB01S
MBlti-MM Mw
Cot-Point*
BC/CO/WO*
0.6/20/2.00
0.8/20/2.00
1.0/20/2.00
1.5/20/2.00
0.6/15/2.00
0.8/15/2.00
1.0/15/2.00
1.5/15/2.00
0.6/ 8/2.00
0.8/ 8/2.00
l.O/ 8/2.00
l.S/ 8/2.00
0.6/ 6/2.00
0.6/ 6/2.00
l.O/ 6/2.00
l.S/ 6/2.00
0.6/20/1.50
0.8/20/1.50
1.0/20/1.50
1.5/20/1.50
0.6/15/1.50
0.8/15/1.50
1.0/15/1.50
1.5/13/1.30
0.6/ 8/1.50
0.8/ 8/1.50
l.O/ 8/1.50
l.S/ 8/1.50
0.6/ 6/1.50
0.6/ 6/1.50
l.O/ 6/1.30
l.S/ 6/1. SO
0.6/20/1.23
0.8/20/1.23
1.0/20/1.23
1.5/20/1.23
to XdM
ac
75%
74%
74%
8«%
73%
75%
75%
88%
76%
76%
76%
8«%
87%
87%
87%
•8%
75%
74%
74%
80%
73%
73%
73%
a.%
76%
76%
76%
M*
87%
87%
87%
88%
73%
74%
74%
•klfioatioa Wiirti*
R»*« of rwLlac*
Sfi
77%
62%
60%
60%
77%
62%
62%
62%
81%
67%
67%
67%
82%
80%
80%
80%
78%
63%
61%
61%
78%
63%
63%
63%
82%
67%
67%
67%
83%
80%
80%
80%
78%
63%
61%
61%
BBS
68%
49%
43%
43%
68%
49%
49%
49%
68%
49%
49%
49%
68%
59%
59%
39%
83%
64%
38%
58%
83%
64%
64%
64%
83%
64%
64%
64%
83%
73%
73%
73%
•3%
(4%
M%
M%
'ili'inrM
13
7
6
6
13
7
7
7
14
8
a
8
13
11
11
11
14
8
7
7
14
8
8
8
IS
9
9
9
X,
12
12
12
14
9
8
8
St»/km
33%
18%
15%
15%
33%
18%
18%
18%
36%
21%
21%
21%
38%
28%
28%
28%
36%
21%
18%
18%
36%
21%
21%
21%
38%
23%
23%
23%
41%
31%
31%
31%
36%
23%
21%
21%
of
ta'm
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
1
1
1
1.
Co
ttafea
3%
0%
0%
0%
3%
0%
0%
0%
3%
0%
0%
0%
3%
0%
0%
0%
3%
0%
0%
0%
3%
0%
0%
0%
3%
0%
0%
0%
3%
0%
0%
0%
3%
3%
3%
3%
railoM B*t«
f o* IT»
ftmmlay
¥feUfllU
9%
0%
0%
0%
9%
0%
0%
0%
9%
0%
0%
0%
9%
0%
0%
0%
9%
0%
0%
0%
9%
0%
0%
0%
9%
0%
0%
0%
9%
0%
0%
0%
9%
9%
9%
9%
r«Uaz* tat*
fa* Mom*!
talttla?
Xfcfciallf
17%
3%
0%
0%
17%
3%
3%
3%
20%
7%
7%
7%
23%
10%
10%
10%
20%
7%
3%
3%
20%
7%
7%
7%
23%
10%
10%
10%
27%
13%
13%
13%
20%
10%
7%
7%
A-3
-------
923
39 Voaiel**
Molti-ASM Mod
Cat-Poiato
•e /eo/MQx
0.6/13/1.23
0.8/15/1.23
1.0/15/1.25
1.5/15/1.25
0.6/ 8/1.25
0.8/ 8/1.25
l.O/ 8/1.23
1.5/ 8/1.25
0.6/ 6/1.25
0.6/ 6/1.25
l.O/ 6/1.25
1.5/ 6/1.25
0.6/20/1.00
0.8/20/1.00
1.0/20/1.00
1.5/20/1.00
0.6/15/1.00
0.8/15/1.00
1.0/13/1.00
1.5/13/1.00
0.6/ 8/1.00
0.8/ 8/1.00
l.O/ 8/1.00
1.5/ 8/1.00
0.6/ 6/1.00
0.6/ 6/1.00
l.O/ 6/1.00
l.S/ 6/1.00
1.6/13/1.72
to XdM
K
84%
75%
75%
75%
84%
76%
76%
76%
84%
87%
87%
87%
94%
82%
80%
80%
94%
82%
82%
82%
99%
82%
82%
82%
98%
94%
94%
94%
73%
*lflo«4
CQ
78%
63%
63%
63%
82%
67%
67%
67%
83%
80%
80%
80%
85%
72%
70%
70%
89%
72%
72%
72%
90%
74%
76%
74%
91%
90%
90%
90%
63%
elo*
02*
83%
64%
44%
64%
93%
44%
44%
44*%
43%
79%
79%
79%
94%
44%
40%
40%
94%
44%
44%
84%
•4%
44%
44%
44%
94%
•4%
94%
94%
44%
of
F.tlt..^
14
9
9
9
IS
10
10
10
16
13
13
13
18
14
13
13
14
14
14
14
19
IS
IS
IS
20
14
19
14
4
railan
ana
36%
23%
23%
23%
38%
26%
26%
26%
41%
33%
33%
33%
46%
36%
33%
33%
44%
34%
36%
34%
49%
38%
38%
34%
51%
44%
44%
44%
21%
of
(all
i
i
i
i
i
i
i
i
i
i
i
i
2
2
2
2
2
2
2
0
2
2
2
2
2
2
2
2
0
la
Bift«
3%
3%
3%
3%
3%
3%
3%
3%
3%
3%
3%
3%
5%
5%
3%
5%
5%
5%
5%
0%
5%
5%
5%
5%
5%
3%
3%
3%
0%
r«ii«M ««*•
f o» n»
PaMiaff
jEafeialAC
9%
9%
9%
9%
9%
9%
9%
9%
9%
9%
9%
9%
18%
18%
18%
18%
18%
18%
18%
0%
18%
18%
18%
19%
18%
18%
18%
19%
0%
railan (Uto
to* NonoJ.
teLttiag
YatAalat
20%
10%
10%
10%
23%
13%
13%
13%
27%
17%
17%
17%
30%
23%
20%
20%
30%
23%
23%
23%
33%
27%
27%
27%
37%
30%
30%
30%
7%
1
XdoBti.«io«fcloa
owkpoiat *•*
A-4
-------
Appendix B
Scatttrplot*
-------
A
I
?:«
o 25
0
M
u
U
•a
u
en
i
M
O
M
M
U
•a
4J
n
-------
o
Ul
Ul
u
8
O I
* I
11
M
O
u
w
u
•o
si
n
41
3
•O i
v ;
4J
O i
4)
U
o>
(ft
V
3
O4
OO 4U
-------
1-4
o
M
M
Ct]
i
04 j
-------
o
o
o
o
u
1
o
to
o
o
in
r-
o
i
U
ul
ei
ut
r- O*
o
I O
u vo
• * I
f»
-------
o
o
o
o
\
m
i
a
I
o
in
I
in
O
I
u
01
ul
o
u
VO * I
• m
CM • M
—
v *J
X U OT
O
O
o
o
e
o
••4
4*
U
1
to
8
5
o
o
I- I
o
o
m
o
o
o
o
u)
-------
o
o
•
i
£
3
CM
in
o
in
iH
o
in
X
CO
at
«wn»
I-
O
U
O
en
r- vo
o <*»
i
U o
• f u
f-l • O
— r-» W
« r- u
X ul
M I
< tl
U O)
O
O
§
o
in
o
o
o
in
o
o
o
o
o
o
•
CM
O
m
o
o
o
in
o
o
-------
I
A
S
(J
1
ocl
4-
HC Emission;
Vehicles -
u
|
m
OB
9k
Ok
*O
*•
d
a
in
0»
4*
••*
*
^
4J
0>
in r*
o
1
CO * M
0 • 0
0 vo U
o on u
U
0 1
1 o< T»
o
o
o
o
I
o
o
o
o
o
o
o
o
a\
o
o
o
oo
o
o
o
r-
M
O
U
U
T3
m i
o
u
M
U
•o
4J
o
0
o
in
o
1
0.
1
U : n
a «
3
in
o
lA
^
-------
A
!
en
«a
X
in
o
o
o
in
I
u
o
u
u
•o
(O
o
o
oo
I -
o
o
0
U
U
U
•a
Cfl
i
0
u
•o
r-l
fl
•o
V
*J
V
M
cu
n
v
3
Ot
o
o
o
VO
o
o
o
m
o
o
o
o
o
CO
o
o
o
04
o
o
o
o
(TV/6) 00
-------
o
o
at
i
I
••4
«o
Ok
O >£>
O O
+• I
.*»
(O • 0
vn cs w
M VO U
u
O I
I CM -O
>i U ttl
M
0
u
u
u
o
o
o
o
o
a
m i
vo
o
O
o
o
o
o
o
o
o
O
o
o
o
u
0
u
M
U
•a
+j
V)
-------
i
01
in in
r»
O
O
X *
<•> (<« U
o • o
o o u
°. N u
O I
I ci -d
< *i
>i U (0
o
o
I
o
o
o
o
in
If) f V
o
o
•
m
o
m
CM
OH
-------
I
A
O
, M
o
o
o
00
a
o
o
a
o
vo
O
o
o
o
o
a
(tm/6) CO
-------
o
o
M
0
u
U
hi
A
i
o o
+• I
d»
W en U
O • O
in o u
vN Is- 14
o . U
I 04 TJ
>t W 0)
O
in
-------
Appendix B:
Estimated Coat of High-T«ch I/M Tasting
-------
5.2.1 General Methodology
EPA's estimates of the costs of high-tech test procedures are driven
by a number of assumptions. Costs in conventional centralized and
decentralized test-and-repair programs were derived using current inspection
costs in I/M programs as they are reported to EPA as the starting point. For
decentralized test-only networks costs are modelled in a manner similar to
centralized programs, since all current test-only programs are centralized,
however, costs are estimated using a range of test volumes and a higher level
of state oversight is assumed since the network is composed of independent
operators and may have a higher number of test sites than in centralized
programs.
Another key assumption is that adding the new tests will increase
inspection costs in programs that are now efficiently designed and operated.
In programs that are not now well designed, current costs are likely to be
higher than necessary and the cost increase less if efficiency improvements
are made simultaneously. In order to perform the high-tech tests new
equipment will have to be acquired and additional inspector time will be
required for some test procedures. The amount of the cost increase will be
determined to a large degree by the costs of acquiring new equipment and the
impact of the longer test on throughput in a high volume operation. Average
test volume in decentralized programs is low enough to easily absorb the
additional test time involved (although at a cost in labor time). Equipment
costs are analyzed in terms of the additional cost to equip each inspection
site (i.e., each inspection lane in centralized inspection networks, and each
licensed inspection station in decentralized networks).
By focusing on the inspection lane or station as the basic unit of
analysis, the resulting cost estimates are equally applicable in large
programs, with many subject vehicles and inspection sites, or small programs,
with few subject vehicles and inspection sites. Previous EPA analyses of
costs in I/M programs have found that the major determinants of inspection
costs are test volume and the level of sophistication of the inspection
equipment. Costs of operating programs were not found to be measurably
affected by the size of the program (for further information the reader may
refer to EPA's report entitled, "I/M Network Type: Effects on Emission
Reductions, Cost, and Convenience"). Figures on inspection volumes at
inspection stations and lanes are available from I/M program operating data.
This information enables the equipment cost per vehicle and the additional
staff cost per vehicle to be calculated for each test procedure.
H-2
-------
The equipment cost figures presented in this paper are based on the
costs of the equipment EPA believes is best suited for high-tech testing.
They are current prices quoted by manufacturers, and do not reflect what the
per unit prices might be if this equipment were purchased in volume. Staff
costs are based on prevailing wage rates for inspectors in both types of
programs as reported in conversations with state I/M program personnel.
Construction costs in centralized programs are based on estimates supplied by
centralized contractors. Other site costs and management overhead in
centralized programs are back calculated from current inspection costs. For
decentralized networks, it is assumed that longer test times could be absorbed
with no increase in sites. The current average volume in decentralized
stations is 1,025 vehicles per year (between 3 and 4 vehicles per day,
depending upon the number of days per year the station is open).
Consequently, increasing the length of the test, to the degree that the new
procedures would, is not expected to impact the number of inspections that can
be performed.
5.2.2 Equipment Need* and Costs
A pressure metering system, composed of a cylinder of nitrogen gas
with a regulator, and hoses connecting the tank to a pressure meter, and to
the vehicle's evaporative system is needed to perform evaporative system
pressure testing. Hardware to interface the metering system with a
computerized analyzer is also needed and is included in the cost estimate.
Purge testing can be performed by adding a flow sensor with a computer
interface, a dynamometer, and a Video Driver's Aid. With the further addition
of a Constant Volume Sampler (CVS) and a flame ionization detector (FID) for
HC analysis, two nondispersive infrared (NDIR) analyzers for CO and carbon
monoxide (C02>, and a chemilumineacent (CD analyzer for NOX, transient
testing can be performed.
The analyzers used for the transient test are laboratory grade
equipment. They are designed to higher accuracy and repeatability
specifications than the NDIR analyzers used to perform the current I/M tests.
Table 5-4 shows the estimated cost of equipment for conducting high-tech
tests. This quality of technology is essential for accurate instantaneous
measurements of low concentration mass emission levels.
H-3
-------
Table 5-4
Equipment Coats for Hew Testa
Test Equipment Price
Pressure Metering System $600
Purge Flow Sensor $500
Dynamometer $45,000
Video Drivers Aid $3,000
Transient CVS 6 Analyzers $95,000
TOTAL $144,100
The figures in Table 5-4 do not include the costs of expendable
materials. Nitrogen gas is used up in performing the pressure test.
Additionally, the FID burns hydrogen fuel. Calibration gases are needed for
each of the analyzers used in the transient test. Because the analyzers used
in the transient test are designed to more stringent specifications than the
analyzers currently used in the field, bi-blends, gaseous mixtures composed of
one interest gas in a diluent (usually nitrogen) are used to calibrate them.
Multi-blend gases, such as are typically used to calibrate current I/M
equipment, are not suitable. Current estimates for expendables are shown in
Table 5-5. The replacement intervals are estimated baaed on the usage rates
observed in the EPA Indiana pilot program and typical inspection volumes as
presented later invthis section. Calculations of per vehicle equipment costs
presented throughout this report include per vehicle costs of these
expendables as well.
Table 5-5
Kxpendablea for New Teata
Replacement Interval
Test Material Coat Centralized Decentralized
Pressure N2 Gaa 930 250 tests 250 teata
Transient H2 Fuel $60 2 months 1000 tests
HC Cal Gaa $€0 2 months 1000 teats
CO Cal Gaa $60 2 months 1000 testa
C02 Cal Gaa $$0 2 months 1000 tests
Staff costs have been found to vary between centralized and
decentralized programs, as does the effect on the number of sites in the
network infrastructure. Therefore, the following sections are devoted to
separate cost analyses for each network type.
H-4
-------
5.2.3 Coat to Upgrade Centralized Network*
5.2.3.1 Basic Aaaumptiona
The starting point in this analysis is the current average per
vehicle inspection cost in centralized programs. A figure of $8.50 was used
based upon data from operating programs. This figure includes the cost of one
or more retests and network oversight costs. The key variables to consider in
estimating the costs in centralized networks are throughput, equipment, and
staff costs. Data on these variables were obtained by contacting program
managers in a number of these programs, and by surveying program contracts and
Requests for Proposal.
Throughput refers to the number of vehicles per hour that can be
tested in a lane. The higher the throughput rate, the greater the number of
vehicles over which costs are spread, and the lower the per vehicle cost. EPA
contacted program managers and consulted the contracts in a number of
centralized programs to determine peak period throughput rates in the
different systems. Rates were as reported in Table 5-6.
Table 5-6
Peak Period Throughput Rates ia Centralised Z/M Programs
Program Vehicles Tested per Hour
Arizona 20
Connecticut 25-30
Illinois 25
Maryland 25-35
Wisconsin 25-30
On the basis of this information, 25 vehicles per hour was assumed to
represent the typical peak period throughput rate or design capacity in
centralized X/M programs. During off-peak hours and days, throughput is lower
since there is not a constant stream of arriving vehicles. Conversations with
individuals in the centralized inspection service industry indicate that
inspectors start at minimum wag* or slightly higher, that by the end of the
first year they earn $5.50 to $€ per hour/ and that they generally stay with
the job for one to three) years. Thus, $6 per hour was used to estimate the
average inspector's hourly wag*.
Estimates of the costs of adding pressure testing, purge testing/ and
transient tailpipe testing were derived by taking the current coats for the
new equipment to perform the new tests, dividing it by the number of
inspections expected to be performed in the lane over a five year period and
H-5
-------
adding it to the current $8.50 per vehicle cost, with a further adjustment foe
the impact of test time on throughput, and thus on the number of sites and
site costs. The same is done to estimate additional personnel costs
associated with adding the new tests, when independent programs were surveyed
to determine the length of a typical contract, it was discovered that
Illinois, Florida, and Minnesota all have five year contracts, Arizona has a
seven year contract, and the program in the State of Washington is operating
under a three year contract, resulting in an average contract length of five
years among the five programs surveyed. Five years was therefore chosen as
the typical contract length.
The number of inspections expected to be performed over the five year
contract period was derived by calculating the total number of hours of lane
operation, estimating the average number of vehicles per lane and multiplying
the two. A lane is assumed to operate for 60 hours a week (lane operation
times were found to vary from 54 to 64 hours per week), 52 weeks a year for
five years for a total of 15,600 hours. Lanes are assumed to have a peak
throughput capacity of 25 vehicles per hour. Modern centralized inspection
networks are designed so that they can accommodate peak demand periods with
all lanes operating at this throughput rate. Networks are usually designed so
that average throughput is 50-65% of peak capacity or 13-15 vehicles per hour.
When operating for 15,600 hours over the life of a contract, a centralized
inspection lane is estimated to perform a total of 195,000 inspections, or
about 39,000 per year.
5.2.3.2 The Kffeet of Changing Throughput
The addition of evaporative system pressure testing to a centralized
program would result in a slight decrease in the throughput capacity. The
addition of purge and transient testing, along with pressure testing, would
result in a further decrease.
Assuming the sane test frequency (i.e., annual or biennial) the
reduced throughput rat* means that the number of lanes needed to test a given
number of vehicles would increase accordingly, as would the size of the
network infrastructure needed to support the test program. The result is an
increase in the cost per vehicle. Actual consumer cost depends on the test
frequency; EPA would encourage states to adopt biennial programs to reduce the
costs and imposition of the program. Less frequent testing only slightly
reduces the emission reduction benefits while cutting test costs almost in
half.
H-6
-------
One way to estimate the cost would be to simulate an actual network
of stations and lanes in a given city. One could attempt to assess land
coats, building coats, staff and equipment costs, costs for all necessary
support systems, and other cost factors. However, this approach would be very
time consuming and would rely on information which is proprietary to the
private contractors that operate the programs and is, therefore, unavailable.
Instead, the cost of the increased number of lanes and stations is derived by
analyzing current costs and subtracting out equipment, direct personnel,
construction, and state agency oversight costs. The remainder is adjusted by
the change in throughput in the new system. Then, new estimates of equipment,
personnel, construction, and oversight costs are added back in to obtain the
estimated total cost.
As discussed previously, the typical high volume station can test 25
vehicles per hour, performing (in most cases) a test consisting of 30 seconds
of high speed preconditioning or testing, followed by 30 seconds of idle
testing. In addition, a short time is spent getting the vehicle into position
and preparing it for testing. This leads to a two to three minute test time
on average, depending upon what short test is performed. EPA recently issued
alternative test procedures for steady-state tests that reduce various
problems associated with those tests, especially false failures, but at a cost
of longer average per test time.
Current costs were estimated by contacting operating program
personnel, equipment vendors and contractors. The most sophisticated
equipment installation (i.e., the equipment for loaded steady-state testing)
was used to estimate current equipment costs.
The cost to acquire and install a single curve dynamometer and an
analyzer in existing networks ia about $40,000 or 21$ per vehicle using the
basic teat volume assumptions. As indicated previously, a staff person is
assumed to earn 36.00 per hour. When this figure is multiplied by 15,600
total contract hours and divided by 195,000 vehicles, direct staff costs are
estimated at 48$ per vehicle. Existing centralized networks typically have
two staff per Ian*. Thus, total staff costs work out to 96$ p«r vehicle.
Total average construction costs are estimated at 9800,000 for a five lane
station, yielding an average p«r vehicle cost of 826. In this analysis a
figure of $1.25 is used to estimate the amount of the state retainer. This
reflects EPA*s best estimate of the per vehicle expense for a good state
quality assurance program in a centralized network. Equipment, staff,
construction, and state costs add up to $3.24 per vehicle. Subtracting this
amount from the current average of $8.50 leaves $5.26 in infrastructure costs
and other overhead expenses including employee benefits and employer taxes as
H-7
-------
shown in Table 5-7. This amount is then factored by the change in the
throughput rate and the equipment, oversight, and staff costs for the new
tests are then added.
Table 5-7
Current Program Costs
Total Cost Less
Increments Per Vehicle Cost Increments
Current $8.50
Equipment $0.21 $8.29
Staff $0.96 $7.33
Construction $0.82 $6.51
State Retainer $1.25 $5.26
5.2.3.2 Costs of New Tests
Most centralized programs use a two position test queue; emission
test are done in one position while emission control devices are checked in
the other, along with other functions such as fee collection. In this type of
system the throughput rate is determined by the length of time required to
perform the longest step in the sequence, not by length of the entire test
sequence. The new tests would likely be performed in a three position test
queue, with one position dedicated to fee collection and other administrative
functions, one to performing the pressure test, and the third to performing
the transient and purge tests. The transient/purge test is a longer test
procedure than the ones currently used in most I/M programs and is the longest
single procedure in the whole inspection process. Thus, it is the determining
factor in lane throughput and will therefore influence the number of test
sites required.
The transient test takes a maxinmm of four minutes to perform. An
additional minute is assumed to prepare the vehicle for testing, for a maximum
total of five minutes. The pressure test would take approximately two
minutes, and could be shortened through such potential strategies as
computerized monitoring of the rate of pressure drop. EPA is in the process
of looking at potential fast-pass and fast-fail strategies, and'preliminary
results suggest that roughly 33% of the vehicles tested could be fast passed
or failed based upon analysis of data gathered during the first 93 seconds of
the IM240 (i.e.. Bag 1) using separate fast-pass and fast-fail cutpoints.
Hence, EPA estimates that the average total test time could be shortened to at
least four minutes per vehicle. This translates into a throughput capacity of
15 vehicles per hour. To accommodate peak demand periods and maintain short
wait times, a design throughput rate of half of capacity is assumed, for a
H-8
-------
typical throughput rate of 7.5 vehicles p«r hour. Assuming the same number of
hours of lane operation as previously, the total number of tests per lane in a
transient lane is estimated to be 117,000 over the five year contract period.
State quality assurance program costs would increase given the
complexity and diversity of the test system; an estimate of an additional 50$
is used here but the amount could vary depending upon the intensity of the
oversight function the state chooses. Staff costs per vehicle are calculated
using the same assumptions for wages and hours of operation as shown in Table
5-7; however, the cost is spread over 117,000 tests over the life of the
contract rather than 195,000. The result is staff costs of 80* per staff per
vehicle. Three staff per lane are assumed to perform the tests. The
additional tasks performed by inspectors in conducting the new tests - i.e.,
disconnecting vapor lines and connecting them to analytical equipment for the
evaporative tests and driving the vehicle through the transient driving cycle
- do not require that inspectors have higher levels of skill than they do
presently. Rather, these tasks can be performed by comparably skilled
individuals trained to these specific tasks. Total staff costs work out to
$2.40 per vehicle. Equipment costs for each test procedure are derived by
taking the equipment coats from Table 5-4 and calculating the costs of five
years worth of expendables using the figures in Table 5-5 and dividing by
117,000. Construction costs for a five lane station are assumed to rise to
$1,000,000. This is due; to the fact that slightly longer lanes may be needed
in order to accommodate test equipment and facilitate faster throughput.
Dividing this figure by 117,000 vehicles per lane yields a per vehicle cost of
$1.71. The resulting costs estimates are shown in Table 5-8. Table 5-8 shows
the result of factoring the figure of 35.26, from Table 5-7, by the change in
the throughput rate and adding in the equipment, staff, construction and state
costs associated with the new test procedures. The figure of $5.26 is
multiplied by 12.5/7.3, i.e., the ratio of the design throughput rate in the
current program to the- design throughput rate in a program conducting pressure
purge and transient testing.
H-9
-------
Table 5-8
Coats to Add Proposed Tests to Centralised Programs
Running Total
Increments Per Vehicle Cost Cost per Vehicle
Adjust for Throughput $5.26 * 12.5/7.5 $9.12
Staff $2.40 $11.52
Construction $1.71 $13.23
Oversight $1.75 $14.98
Pressure Test $0.13 $15.11
Purge Test $0.41 $15.52
Transient Test $0.87 $16.40
Thus, the cost of adding the new tests to centralized networks is
found to be about double the current average cost. The cost of centralized
test systems has been dropping in the past few years as a result of
competitive pressures and efficiency improvements. These factors may drive
down the costs of the new tests as well, especially as they relate to
equipment costs. Given that conservative assumptions were made regarding
equipment costs of $144,000 per lane, and low throughput rates, the cost
estimate presented here can be fairly viewed as a worst case assumption. As
discussed earlier, the important issue is the quality of the test, not the
frequency/ so doing these tests on a biennial basis would offset the increased
per test cost.
5.2.4 Coat to Upgrade Decentralized Programa
5.2.4.1 Baaie Assumptions
The methodology used to estimate costs in decentralized programs is
similar to that described above for centralized programs. Equipment and labor
costs are key variables aa they were in determining costs for centralized
programs. However, estimates of costs for decentralized programs presented
here do not include estimates of land costs and overhead. While inspections
in decentralized programs are generally conducted in pre-existing facilities
rather than newly built ones, there are nonetheless a variety of overhead
expenses as well aa opportunity costs associated with making space available
for inspections in a facility that provides a number of other services as
well. Data on these costs are not available and they cannot be deduced from
reported inspection fees since, in most programs, fees are capped by law and,
hence, do not reflect the actual cost of providing an inspection.
H-10
-------
Total test volume rather than throughput and test time are the
critical factors affecting cost in decentralized programs. Licensed
inspection stations at present only perform, on the average, about 1,025
inspections per year, as shown in Table 5-9 (note that this number is a
station-weighted average). Test volumes among stations in a single program
can vary widely as shown in Section 7.0. It should also be noted that all
decentralized programs in enhanced I/M areas, except for California, Virginia,
and Colorado (which tests vehicles five years old and newer biennially, and
vehicles older than five years annually) are annual programs. In this
analysis the effect on per vehicle costs of switching from an annual
inspection frequency to biennial, as well the effect of varying inspection
volume, will be examined.
Table 5-9
Inspection Volumes in Licensed Inspection Stations
Program Vehicles per Year Vehicles per Station
California 6,180,093 799
Colorado 1,655,897 1,104
Dallas/Ft. Worth 1,948,333 1,624
Bl Paso 278,540 1,161
Georgia 1,118,448 1,729
Houston 1,482,349 1,348
Louisiana 145,175 1,037
Massachusetts 3,700,000 1,321
Nevada 523,098 1,260
New Hampshire 137,137 564
New York 4,605,158 1,071
Pennsylvania 3,202,450 834
Rhode Island 650,000 684
Virginia 481,305 1,301
Weighted Average 1,025
Annual tests of 1,025 vehicles per station is equivalent to between
three and four inspections per day depending upon the number of days per week
the facility is open and inspections are available. This is far below the 75
inspections per day projected in a multi-position high volume lane with three
inspectors conducting high-tech tests, and significantly below the 16
inspections per day that could be don* in a single position inspection bay
with only one inspector (the derivation of this figure is detailed below).
Two conclusions can be drawn from this. The first is that the additional time
requirements of the new tests will not force a reduction in the total number
of inspections that most stations can perform. The second is that, because
costs are spread over a smaller number of vehicles than in the case of high-
H-ll
-------
volume, centralized stations, the coat per vehicle for the new teats will be
larger in this type of inspection network.
The higher costs for high-tech testing equipment have prompted
questions of whether all current inspection stations would choose to stay in
the inspection business with the implementation of an enhanced program, and
how high a drop-out rate programs would experience if some did not. EPA knows
of no data or reasonable assumptions by which a station drop-out rate could be
reliably estimated. In this analysis inspection costs for high-tech testing
are estimated for three scenarios: one where all stations remain in the
inspection business, one where 50% of the stations drop out, and one where
enough stations drop out such that those that remain are operating at maximum
possible volume assuming that each has one inspection bay which has not been
improved for high throughput and one inspector performing all parts of the
inspection. In all three scenarios a biennial inspection frequency is
assumed.
The current average test fee for vehicle inspection in decentralized
programs is about $17.70 (again, the derivation of this figure can be found in
EPA's technical information document, "I/M Network Type: Effects on Emission
Reductions, Cost, and Convenience"). Note that this figure may substantially
underestimate actual costs since most states limit the inspection fee that a
station may charge. In many cases, the actual fee is likely to be below cost;
stations presumably obtain sufficient revenue to stay in business by providing
other services, which may include repair. It should also be noted that the
intensity of the inspection and the sophistication and cost of the analyzer
vary significantly among programs. Average inspection costs and revenues by
program, taking these factors into account, are estimated in Section 7.4.1.
The costs for adding high-tech tests are derived by estimating the
per vehicle costs of the key components: labor; equipment, including
expendables; and support, i.e., service contracts and annual updates. Per
vehicle costs are estimated by deriving total costs for each component and
dividing by the number of vehicle inspections expected to be performed in a
year, again, taking into account variations in inspection volumes and changes
in frequency. Equipment: costs are spread over the useful life of the
equipment. While a piece of equipment's useful life can vary considerably in
actual practice, a five year equipment life is assumed.
While large businesses, such as dealerships, may be able to afford to
purchase current analyzer equipment outright, the smaller gas stations and
garages typically have to finance these purchases (although in some cases they
may lease equipment). The higher cost of the equipment needed to perform
H-12
-------
purge and transient testing ($144,000 for the dynamometer, CVS, analyzers,
etc., as opposed to $12,000 to $15,000 for the most sophisticated of the
current NDXR-based analyzers) makes it even more likely that these purchases
will have to be financed for most inspection stations. Equipment costs are
amortized over five years at 12% interest in the analysis in this report.
Program personnel in decentralized programs were contacted to
determine inspector wage rates. In many cases, inspectors are professional
mechanics earning about $25 per hour. However, most states do not require
inspectors to be mechanics, and inspections may be performed by less skilled
individuals who typically earn $6 or $7 per hour. The prevalence of different
wage rates among inspectors is unknown. Therefore, EPA assumed an average
wage of $15 per hour for this analysis. An overhead rate of 40% is assumed,
for a total labor cost of $21 an hour.
5.2.4.3 Coat Components and Scenarios
The full test, including data entry on the computer, preparing the
vehicle for the different steps in the test procedure and conducting them, is
estimated to take 30 minutes with only one inspector performing all tasks in a
repair bay that is not configured specifically for inspection throughput.
With labor costs at $21 per hour, as described above, this works out to $11.50
per vehicle. Equipment costa arc taken from Table 5-4 and are amortized over
a five year period at 12 percent annual interest (changing the assumed
interest rate does not significantly affect the total per vehicle cost). This
brings the total cost for the equipment package over the five year period to
$192,325. These costs arc divided by five years worth of inspections. The
costs of expendables from Table 5-5 are added in according to the usage rates
assumed for decentralized programs. Two other expenses typically encountered
in decentralized programs arc service contracts and software updates. Based
on information from states, service contracts are estimated at $200 per month
and annual software updates are assumed to cost $1,500.
Per vehicle costa arc estimated for three scenarios, biennial testing
is assumed in all three. la the first, all stations remain in the inspection
program. la the second> 50 percent of the stations drop out of the program,
and in third there are only the minimum number of stations in the program to
enable each to inspect at full volume with one inspector performing all parts
of the inspection and a service station bay that has not been improved for
high throughput.
In the first scenario, the switch to biennial would mean that annual
volume is cut in half, or 513 vehicles per year. In the second scenario the
H-13
-------
SO percent reduction in the number of stations brings the annual inspection
volume back to 1,025. In the fourth scenario, it is assumed that each station
inspects at maximum capacity, i.e., one vehicle every thirty minutes, and that
an inspector is available 50 hours per week. This results in an annual volume
of 5,200 vehicles.
Table 5-10
Cost* to Conduct High-Tech Testing in Decentralized Program*
Scenario Annual Volume Cost per vehicle
No Drop-out 513 $106
50% Drop-out 1,025 $58
72% Drop-out 5,200 $32
(Maximum volume)
Note that while reducing inspection frequency to biennial cuts
motorists' costs in centralized programs, in decentralized programs such cost
reductions are only achieved by reducing opportunities for stations to
participate. In the scenario in which 50 percent of the stations drop out and
testing is biennial, annual station volume is the same as if testing were
annual and no stations dropped out. Hence, the estimated per vehicle cost in
a biennial program with a 50 percent station drop-out rate is the sane as
would be derived for an annual program with no stations dropping out.
Reducing inspection frequency to biennial, while maintaining the same number
of stations, has the effect of almost doubling the per vehicle cost since
operating costs are spread over half as many vehicles. Note also that the per
vehicle cost far exceeds the per vehicle cost in centralized programs except
in the scenario where 72 percent of the stations drop-out.
5.3 Costs) of four-Mode, Purge and Pressure Testing
It has been proposed that a series of simpler, loaded mode and other
steady-state tests would provide equivalent emission reductions to the IM240
at a lower cost. The emission reduction potential of this approach is
currently being evaluated at EPA's test lane in Phoenix, Arizona. The
information needed to do a cost analysis can be approximated at this time
based upon the test process.
The test procedure being evaluated is a series of emission tests
referred to as the four-mode test: A 40 second 5015 mode (IS mph at a load
equivalent to ETN / 250), a 40 second 2525 mode (25 mph at load equivalent to
ETW / 300), a 40 second mode at 50 mph and normal road load, and a 40 second
idle mode. EPA anticipates a 30-60 second preconditioning mode would be
H-14
-------
needed to insure proper warm-up and canister purge down. Allowing also for
necessary tin* to transition between test modes (5-10 seconds), the four-mode
test would require a total of approximately four minutes. As with the IM240-
based test scenario, purge testing is assumed to occur simultaneously with the
tailpipe test and pressure testing would be done separately. It should be
noted, however, that some vehicles may not purge during this test and may
require a short transient retest to activate purge.
5.3.1 equipment and Expendables
The equipment used for the four-mode test is simpler than for the
IM240 test. The dynamometer may not need inertia weights, and a raw gas
analyzer, like the ones used in the current I/M testa, is upgraded with a NOx
analyzer and an anemometer, to enable mass concentration calculations, for
this test. The equipment for the purge and pressure test are the sane as
described previously. The estimated costs are shown in Table 5-11.
Table 5-11
equipment and Cost* foe the ASM Test
Pressure System $600
Flow Sensor $500
Dynamometer $20,000
Anemometer $2,000
BAR90 w/NOx Analyzer $16,900
Total $40,000
Expendables for this test are nitrogen gas for the pressure teat and
calibration gases for the analyzer. The cost of nitrogen gaa ia the same as
in the previous analysis oa XM240 coata (the pressure teat procedure ia the
same regardless of the type of tailpipe test used). Current calibration gases
are multi-blends consisting of propane, CO, and CO2. A coat of 945 per bottle
is used here. In thia analysis, it ia assumed that multi-blend gases that
include NO will b» available at the same cost. Alternatively, one could
assume that two bottle* of calibration gaa, one current standard multi-blend
and a bottle of NO will be needed, however, the additional coat per teat ia
insignificant (leas than 5$, even in a low volume situation).
5.3.2 Centralized Vrogxaaa
The total teat time per vehicle would be about 11 minutea, including
administrative processing in an efficiently run testing lane. In a multi-
position lane the throughput would be governed by teat time at the longest
position, which would be four minutea. Thia translates into a peak throughput
H-15
-------
rate of 15 vehicles per hour and, using the standard design criteria for
centralized programs described earlier, an average throughput of 7.5 vehicles
per hour. Using the lane operation assumptions detailed earlier, this
translates into 23,400 vehicles per lane per year and 117,000 vehicles over an
assumed five year contract period. Three staff per lane would be needed to
perform the entire test sequence including inputting vehicle identification
information, conducting the tests and presenting and explaining the results to
the motorist.
The per vehicle cost of the four-mode test in centralized programs is
estimated by the same methodology as was used to estimate IM240 costs.
Current costs for test equipment, staff, state oversight, and construction are
subtracted from the current average per vehicle cost, this amount is factored
by the change in throughput, and estimated costs for equipment, staff,
construction, and state oversight in a four-mode test program are added to
obtain an estimated total cost.
Table 5-12
Coats to Add Proposed Tests to Centralised Programs
Running Total
Increments Per Vehicle Cost Cost per Vehicle
Adjust for Throughput S5.26 * 12.5/7.5 $9.12
Staff $2.40 $11.52
Construction $1.71 $13.23
Oversight $1.75 $14.98
Pressure Test $0.13 $15.11
Purge Test $0.18 $15.29
Four-mode Test $0.35 $15.64
5.3.3 Decentrallied Programs
The same methodology used to estimate costs of IM240 testing is used
here. Most assumptions are unchanged. Total test time is thirty minutes,
equipment is amortized over a five year period. Two parameters are changed in
this analysis: equipment costs total $40,000 instead of $144,100, and state
costs include a coat for state mass emission testing.
Table 5-13
Cost* to Conduct roue-Mod* Testing in Decentralized Programs
Scenario Annual Volume Cost per Vehicle
No Drop-out 513 $51
50% Drop-out 1,025 $31
72% Drop-out 5,200 $25
H-16
-------
Appendix I:
ASM and IM240 Credits for State Implementation
Plans With MOBILES Runs
-------
en
j
14
Hi
^5
n
ll
nt
II
H «
9*
....3"
«««»(<*SjN
OOOXXXXZ'tt
... $
<4> « W -4
« A 0> u
o.
333 : -
111 .14 t
353 3fl
>t >• >l •• O I •• •• O
-IH-I • al •• a pi •
a a s •* • 41 • I S »4
3>iii«!9N!
:::,iis*4si
XXX i c2 4i'
Ck tlH r- • ~ -
O O O
H
•
Cb •
o 004 u a
MMM lj 0 4
I I I • 41 g
Ki-liH-H O I
r»r-i-ri-* H
IK at« M •. a
•**•** u u 4
!:s:
• u
539 •"« * |
a sssf-a::
.. i^^f
i
H
b
O
O fH * iH
O 10 O M
O in M m
333
' f
^::0:
0 X X X
w z ••
[»$ j
M O
0 O
U O
0 M
fei
bi
bi
b
O «0 CM r4 O O O *• *
k
o MM
^5
« Z n r» <
OOrffHOOOOl
B
o i-ii-i
H
-------
s*»
134
3821
^0.0.41,
•M * •*
0^93
S Y8:
<3\
0 O
410
001
« „
I *
I o m A 1-4
Ka o in eo n
IA o m -4 rt
M
<0 00
O N 0«
0 OO
o « «
o r^ rf
o ww
o 0 O
A «
f-i A
b
I H
n r« 6 two wmr-vwMwn
a, ji |S ° "" 5 *
rt rtOO OO«i-l
• n I co «»M«MWr-»
J) Fl MO
O
o tf>Mm
on I CM
j > I r- •»»
.... § H« --
g^ s B0 MN
-H 3
iSlj
fe
o » »
o r- r-
o o o
> I o 4O w m
IA d N i4 d e d d » «
M -<
I SL.H
|«0 _r*MOOOO»-«
u w
OO
41 O
O N
B
U
A£
A OA«OA«»«A
o iiooooono
I
i
r,
I -
i
I
H
-------
CO
vo
«MMMMOM«0
MMOOOO»M
O
I O
!«
nu
M
• in w
i^d
r:
]!i
I S „,.
u h° •»«•«
» |M° ""*'
> I r» MM
g to o MM
H WJ O MM
Ik
ik
Iki
n « r-
* r- r»
o « o
M A
N N
« o in « <<» «
k!
N I « «MMin»»rtA
• O O«MNMOr-»
« X « « -•« »<
r^v,IOOr-l Am o n M CD
• &B> H K«IM~XIO ooiMMono
95 5 -o o •••
uuiq MIOOMMOOOOVM
T Qi O. U J M — M
MZ -
» IM r- •
O O «
§ u B i «w
i t o> 1 <» o o> w o> 01 » IM r-
•H
-------
en
en
vo
>t
I
o£ -U
-j i *
*••• H I O t» r» »H
b g to o o I o mm Of
innoKoowA «n
r- r- 3 IAO o o -!'* HW
•••• • IAO WNHOOOA«
»J M tc!*r:
s.*J*a9B P"* 5—
5W«'!i
3 u u
5 ;
=s
I S«o.
^.&:°.
P°.—
-------
en
at
tMt-tOOOO-Oi-l
O rt 01 t^ (H I n
I hs °°
§1 * p° ""•
I
£* § [•? «3
*" 9 lei o o o
in» H
-d
> «•
^ > I O M^
!•? a H:° **
. r- 3|«o o o
>M M
fe
n:
4J Ol
134
U -H 41
n
4J »
34
•H 41
fi
H4 • •
OXX!
'iis; i <»;
iir(-"-
in m o w •* 0 o i^ <•
O w r* i^ ^ w »-4 »-4 m
to <• A « CKW «l~
04>4OOOOr>i-l
I «
I o» o«M• o n <•
M0 **'*oooeg'<
... 1
XkCk 4>>
«4X -H
0*83
M ' P
S4J R
I 01
•HWI
• 4» «(
s a5-
5-
2§
1 2 » «
n •i.'^ ."I*"J"!®* .
po^^oooon-
• si
I
K, ;. 1 .-
I li :.«
o m »-»«n o « o
*§
Q«
II
13
~a
in »
i>d
intn
PT>
I «<•
r- o
mn
n
i.i
is*
i ^
I O » » n r4
M o r- r> « n
Ud o d ^t-i
V. I M
> I o <• <• CD o>
nQMOinin «t*
i- 3 latd d d rf-«
M p
fe
n nwnx«««o
o r>N<»«inoo«
O «Nf«OOO««
i
A n t^ N *H o
*4 KO OO O <• •
I ""
2 LS 53S55SS
,9 B° «^«««»°»
ill-
} Ot 0> .
S^j'
sMr§b. i o«»»««,«.,»
• i'SS r'T2 p
»x .-5
5 i1
on
li I
in
I
H
-------
ro
vo
fi-t.-l~ % Uo
o n 01 * n «?-
o « I n •»<0o«or-w««
U I « g « o o « c» n n o
4J*q p • u
qiti mo M^ooooni-i
+ \* a H
o o
4J O
o N
t~
Typ«:
s ii
^3
o +
I
O mr4(n o •* o
n m
t» i»
o
• M
k!
bi
fe
b
H
O l-> -H
O m m
o o O
o o » w » N o n «o
O n^oooootn
• I S r-«eoin»nr4r.
d H wo o in •* M ^ o » «
Mi3o9M -i"4
MX -4 •
o* i a y
Mil O
s *• 5^ i «*•
4 I O> I « O » »i « « in
-------
PI *O O * «» *
CD * I 'H
•«
OI
:s.
288S
g i i i i
1 b • U UOI • d O I «IO>..
M 3 a! Q.r-i o o NO H>IHM> •» • • •
Si 3rfg«8~"- g-HH-H 6>Q H H U H g «* 41 41
o >I S u M--4 o o g S3 S S H v»fl-J-}4
ttfs-Hii"!*1
SS3333I333*
01 » h J * * O H * H M
i >l •• 0 • •• •• O
osi^lhri3
t> >••»> a
4J 41 41 0 13 • O 04 0
:ssii..«"
HHH M-4 » « .
-^ *••
I
o n«f
o « 4
AMNNNOV^
^F4oooomF4
k
H
ki
!| M3?"
2x • «S• • *
09 r-co r-«•«•«>• r»
O n»e
O N>4<
AZNWNmp|iifAr«
^•XOONNNOVt
o a-*** o o o o •••*
-------
n
CM
H
•H
M
oiwninor-ao
£
,§
I w
I O n A i* m onr i
ho o « CD M •» r-1
IAO mr-tm o«c
l-l IN
SS> I A
> I w o o
a koo c««i
ai § Ud I O rtr-l
m n a fa o w w
. . Q I . . . .
r- r- «3 MI o o o
CD N M
£2 > I s .-
u« »
» at
H n
mo
nn
133
41 -H 41
I~ n |H OT N W A «
I W
Jl H (WO M » M M0« O O W
««!o3 B° NMOOOO«rt
ox>.£
s: , .5
bUUi* 1 A X »4 M
, .
1 A X »4 MO no w o r«
f« H S f O P» N «• O » «•
\ ° ' M ........
mo Oi-iiHoooof*i<-i
M -> rt
..93 , 8
• 41 M I fl 41
• I 0> I r-l O »« W A A W A O
^•MU IWW «in«Hr-INOini-t
• •"JH G •••
s a" |2°g'«>°°oo3-
l-)Ch»4l.
I ? »»
Q koo -<-i
§•8 P° NM
I"
c 41
n w
CD (<•
k
o » »
o r- r-
O O O
> I O w w «g>
innQlwoinin » M
r>r>3|Aooo *•*
> I n r«c>«A
U wo n*«««Aoom
JA nMHOooo*
ill s
OHM:
I S
IAO i-i -idddd w «
I Aznnw
Ht^*!*" ..-.-. :
kjoOi-iooooonH
5j| I
M3fi^ i wS
J,^R l.S2SS32?S^S
34liQ I • • b •
5aa ICJO^OOOOOMM
Sn 3
00 •
S8
Utitttt
.
k
•HOOOOONr*
W
o n A iH
o m IB IN
i*
§•3
H 9
t> O
CD N
U V."x
1-5-
o MX;
^J i* ••
3221!
^ o»e> 41.
« * --1
I^Sjg
' Sjf
2a!
si
i ^
H I O -•-
S k«o r-r-
3 IAO o o
I ™
> I O -IrH
O Ko O in n
3 Ud do
CO
I
H
4
P° Hrt<'0002'<
T
I Axon«nn»o-4
|S —rt ^
|«OW»^AA»«A
Ud riddddddd
H d »^
35 . «p
;. -8 -J:
-------
CM
H
•r»
M
a
3
rt p- » p- It CM ••<
o « rt I f Ml-t
Q KO O <-l •*
Q I ' • • •
8 IAO CMCM
IS A A
S MO ««
9 |jd dd
;« i LI ss
'* Q I * * * '
•r- 3 IAO oo
i CM M
•-•A
CM CM
r4t~
H 9
JiJ
m -io
o^S
41 ..
I S »HH ««(••
d S Ue V 14 fl M d O <• «
44 •• 8 Uo HHadddiA^
»• • • 5 3 P "
"J
^
I A -4
|AX«niiini>««OM
jOJOO^MOOOOOCMrt
, J
I A oonOA
ff^atriit0.':"
IAO dooooooo
1 QiO. 41
I to o o cow
.... § H° "^ "•
§1 8 12° NN 2*
S3
H I «n
— 3 N I O AA COM
b o|«o I O rf t««
55
I
H
-------
m
(M
« m
p-
W
•' MV.^
• r- • • m »
o M v o a .0»
-|o**S
CMO OM«or-1
cnnoooow><
I «
I eg o nonww <»r>
MO «n«N
j X lai o m IH m
0 in 1 at
J > I « o o
e • 9 ki o N N
8] B
-t i -^
-~4 H I O 00
b Q K» O CD 00
— Q I . . • •
3 mo o o
mte |)-<
in m
r- r-
(D ««««n
to ««o"ooo^
IS m
S4» M n 41
I o> I <-> 0 « »P- « •» « 01 o
-l-uG M« • « •(N.-.IOO A«
3««2 • • c.
qi23 loio N>-I o o o o « <-•
* Ft d H
i
o n a* M n » r-
m a M vr- r-
o in *4 m o •
c J
si
> I r- •»•
* liO
I
H
Sw w
n in
i *
I n »or-Hi-i««<»
MO r~io«»«oo«
mo lortt-iooon^
H •»
§M A m o *> irt
« N is m o n
N I « mmom*inf-«
Q mo nr«oooOr4"<
5 o3 M n
J 9 Z 4
'33L i sL,.«««MO
^§••98 hrS"*^"?"*
1 " ** 9 B°S.N« •
• 4 >9 3 o • q q 3 s
• • • • Si 3 is H ^ ^ <
SH > .4 f" ' -- ^ - ' '
\ > \
o o o + o
-------
CO
O\
<3\
CM
CM
H
•r|
M
I
i *
>§
--8
si i
13
** s
«>••»
• m w
>« • •
i» o
r:
B«
4J W
134
^U
o J S
o n A >4 n «r-
o in o r< •* r* r-
o « •*!•! o « o
o nn
d do
I n on«o«r-mo
M n
f|
i N I CD r-<-iainA«A«
41 H laid nrtdddddei
w 0 0 O 3 M n
R5j-
M 9
! J-
I * 4*
bd^^dddoS5
JAO NHOOOOAr4
0 O
O
I
O «i-* m
o
I O
r- >
13
n
§A A
<0«
t* O
co n
> i o o o
mm Q V* O in«
" " Q LJ * * *
r* t* M W* o o o
« C4 M
r"
^-j i«
r- o > I
«
iu
sli
O O O O V *4
-5i3
I 3 5 01 ik *< a H — (M
X
•4
0
•M!
i«
b Oi Ik
MX •
O^J
S Y8> I AOOW^A^n«0
3 3^^ |^d^-;dddo«-;
I
H
-------
Appendix J:
Emissions Analyzer Price Information from Horiba
-------
HORIBA
HORIBA INSTRUMENTS INCORPORATED
3901 Varsity Drive. Ann Arbor, Michigan 48108
Telepnone-1(800) 3 HORIBA. In Mich 1(800) 624-0899 or (313) 973-2171
Fax (313)973-7868
April 7, 1993
Environmental Protection Agency
2565 Plymouth Road
Ann Arbor, Ml 48105
Attn: Mr. Bill Pidgeon
Re: IM 240 Analyzer Information
Dear Mr. Pidgeon:
This letter is a follow-up to prior discussions we've had regarding the list price of IM 240
Analyzer Systems. We would like to thank you for the opportunity of discussing our equipment and
market with you and your staff.
We would like to make a clarification in reference to the IM 240 pricing. Horiba is actively
working with six of the seven centralized contractors. Four of these contractors currently have IM 240
analyzers installed. The current list price for the Analyzer/CVS System is $75.515. It should be noted
that this price does not include a blower or external sample line. As you can understand, this is a
"single unit price" and does not reflect discounting for quantity orders. For long-term pricing
considerations, it should be recognized that we also anticipate price reductions following improvements
in manufacturing efficiencies.
Horiba's analytical system can be supplied with other options, such as; a driver's aid, purge
and pressure equipment, data collection and processing capabilities.
IM 240 Analyzer System
HC - FID
CO • NOIR
CO, - NOIR
NOx - Chemiluminescent
CVS - 500-700 CFM
Total: $75,515
-------
EPA Page 2
Mr. Bill Pidgeon
We feel that our forte' is in the analytical and sample handling portion of the testing lane. For
this reason, we are providing you with analytical system pricing only. Most of our customers have
sourced or built the other components themselves.
If you should have any additional comments or questions, please feel free to contact me at 1-
800-3HORIBA.
Sincerely, /
/ " ^ ,;•:/• V
Kenneth W. Thomas
Marketing Manager,
IM Systems
KWT/pm
cc: Neal Harvey
Andy Marko
kt041 .Itr
-------
Appendix K:
Centrifugal Blower Price Quotation
from Combined Fluid Products Company
-------
COMBINED
FLUID
PRODUCTS
COMPANY
TO: Environmental Protection Agency
2565 Plymouth Road
Ann Arbor, Michigan 48105
QUOTATION
ISSUED FROM
G 80S Oakwood Rd., Late Zurich, IL 60047
Phone (708) 5404054 FAX (708) 540-0513
G 125 N. Executive Dr., Brookfield, Wl 53005
Phone (414) 258-7770 FAX (414) 821-1492
G P-O. Box 216, 24 S. Given St., Brownatounj, IN 46112
Phone (317) 852-3961 FAX (317) 852-2337
% 5025 Venture Or., Ann Arbor, Ml 48108
Phone (313) 930-2024 FAX (313) 747-7040
Please Note:
A/R Dept. is located in Lake Zurich. Illinois
ATTENTION:
Mr. Dan Sampson
EFFECTIVE DATE: January 27, 1993
REFERENCE:
QUOTATION NO.:
AA408
EXPIRATION DATE: February 27, 1993
in Compliance With Your Request. We Are Pleased to Quote You As Follows:
QUANTITY
DESCRIPTION
PRICE
1-10
Paxton Centrifugal Blower Model RM-87, including: 10 HP
electric motor running at 3*600 RPM on 230/460 volt
vacuum* three-phase/ 60 Hz/ TEFC motor.
- Inlet Filter/Silencer
$3,320.00
Per Unit Net
11-100
$2,656.00
Per Unit Net
DELIVERY Six to eight weaka P.O.B. Santa Monica* California (Delivery Subject to Prior Saiei
TERMS. Net 30 Days. Subject To Credit Approval
>. Corrunker* Sales Engineer
This quotation subject to standard terms and conditions o> sale aa stated on reverse trie.
-------
Appendix L:
Average IM240 Test Time Utilizing
Preliminary Fast-Pass and Fast-Fail Algorithms
-------
Average IM240 Test Time Utilizing
Preliminary Fast-Pass and Fast-Fail Algorithms
The objective of this analysis was to estimate the average IM240 test time
using algorithms that allow vehicles with very low emissions to fast-pass and
vehicles with very high emissions to fast-fail. This reduces the average time
required for the IM240, allowing higher throughput, which reduces the number
of inspection lanes required. The reduced number of lanes lowers equipment
and personnel costs, having the potential to significantly improve the cost
effectiveness of the I/M program.
This analysis describes the fast-pass and fast-fail algorithms used to
estimate the average IM240 test time. The results are preliminary,
representing what could be achieved in time to comply with the court ordered
deadline for this rulemaking. Developing these algorithms requires using
second-by-second data for HC, CO, NOx, and purge, which is very time
consuming, given the huge amount of data per vehicle.
The ideal fast-pass/fast-fail algorithm consists of two continuous
functions. One function represents emission levels at each second of the
IM240 that reliably predict a passing result while the other function
represents emission levels that reliably predict a failing result. Because
this requires evaluating the results at each second of the test for each of
the vehicles, we determined that this could not be achieved under the time
constraint. Instead, we evaluated nine segments (modes) of the IM240, which
significantly reduces the burden, but gives a less than optimal result.
So, additional fast-pass and fast-fail algorithms will be evaluated in the
future, and additional vehicles will be available for those analyses, so these
results should be regarded as preliminary. For example, very low emitters or
extremely high emitters can be fast-passed or fast-failed early in the IM240
cycle, while vehicles near the certification emission levels will require more
time to accurately predict a passing or failing result. The emission
reduction benefits, obtained from repairing vehicles whose emission levels are
slightly dirtier than their certification standards, are not very cost
effective. Similarly, it also may not be cost effective to run the full IM240
as required to accurately distinguish marginal emitters that pass the full
IM240 from marginal emitters that fail. This can be evaluated by comparing
IDRs, failure rates, and error of commission rates for each second of the
IM240 to determine the best tradeoff.
Another consideration is the IM240 reversed. The IM240 was designed as a
two-mode test. The second mode includes the maximum speed of 56.7 mph. The
L-2
-------
IM240-reversed starts with this high speed mode, then is followed by the low
speed mode. This may further reduce the average test time required to
distinguish malfunctioning cars from properly functioning cars. It should be
especially helpful in rapidly determining whether the purge system is
performing adequately.
The algorithm used in this analysis was comparatively crude due to time
and data handling constraints. Several discrete modes of the IM240 were
selected for determining passing and failing emission levels. These modes
were selected to avoid ending the test during an acceleration or deceleration
and to provide a reasonable duration for each of the nine modes. The average
IM240 test time was calculated as the average of the selected mode times
weighted by the number of vehicles passing or failing at each mode. A more
detailed description of the data and methodology used as well as the results
are included in the following sections.
The database used for this analysis conformed to the model I/M program, so
it was limited to 1986 and newer vehicles with second-by-second IM240 results
- 494 vehicles. These vehicles were tested between June 4, 1992 and August 4,
1992. Data were only used if the composite results calculated from the
second-by-second data had passed EPA's quality control measures. Due to the
volume of second-by-second data and the time constraints involved, the second-
by-second data were not QC'd separately.
The following nine modes were selected for pass/fail determinations:
Modes For Evaluating Fast-Pass And Fast-Fail
Mode
(#)
1
2
3
4
5
6
7
8
9
IM240Mode
(sees.)
0-34
0-60
0-74
0-93
0-113
0-154
0-173
0-206
0-239
IM240 Speed
@ End of Mode
(mph)
22.6
30.4
29.8
0.0
27.2
26.0
47.2
51.6
0.0
To determine the passing and failing emission levels for each mode/ the
sample was divided into passing and failing vehicles. The pass/fail
determination was made based on the "two ways to pass" criteria with 0.8 g/mi
L-3
-------
HC, 15.0 g/mi CO and 2.0 g/mi NOx as composite IM240 outpoints and, 0.5 g/mi
HC and 12.0 g/mi CO bag 2 outpoints. One liter of purge volume was used as
the outpoint for purge flow. These criteria are illustrated below.
Pass/Fail Decisions Based On Two-Ways-To-Pass-Criteria
Decision
IM240
HC
g/mi
Fail
Fail
Fail
>
5
5
0.8
0.8
0.8
IM240
CO
g/mi
515.0
>15.0
5 15.0
Bag 2
HC
g/mi
> 0.5
50.5
50.5
Bag 2
CO
g/mi
5 12.0
>12.0
5 12.0
IM240
NOx
g/mi
5
5
>
2.0
2.0
2.0
Purge
liters
51.0
51.0
51.0
Comments
Must fail HC
Composite &
on both
Bag 2 to
fail
Must fail CO on both
Composite &
Bag 2 to
fail
Only 1 way to Pass:
Composite NOx 5 2.0
Fail
Pass
Pass
Pass
Pass
Pass
Pass
Pass
5
5
>
5
>
5
5
5
0.8
0.8
0.8
0.8
0.8
0.8
0.8
0.8
515.0
5 15.0
>15.0
515.0
5 15.0
5 15.0
>15.0
5 15.0
50.5
50.5
50.5
> 0.5
50.5
> 0.5
50.5
50.5
512.0
5 12.0
5 12.0
>12.0
5 12.0
5 12.0
512.0
>12.0
5
5
5
5
5
5
5
5
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
>1.0
51.0
51.0
51.0
51.0
51.0
51.0
51.0
pass.
to
The minimum emission levels and maximum purge volume for failing vehicles
at each mode were used as fast-pass outpoints. Conversely, the maximum
emission levels for passing vehicles at each mode were used as fast-fail
outpoints. Vehicles were not fast-failed based on purge results since many
vehicles purge late in the IM240 cycle. As mentioned, the IM240-reversed may
help rapidly determine if the purge system is functioning adequately.
The modal outpoint levels, the number of vehicles fast-passing or fast-
failing at each mode and the average IM240 test time as a result of the
application of this fast-pass/fast-fail algorithm are displayed in the
following table.
L-4
-------
Number
Fast-pass Outpoints of
Time Purge
2
3
4
5
6
7
8
9
0-34
0-60
0-74
0-93
0-113
0-154
0-173
0-206
0-239
<0.479/1. 02/0.99
<0.487/0.89/0.99
>0.3
<0.429/0.929/0.90
>0.3
<0.377/0.921/0.84
>0.4
<0.460/0.932/0.89
>0.5
<0.567/1.088/0.96
>0.6
<0.697/3.52/1.33
>0.7
<0.916/14.99/1.77
>0.8
50.805/15.05/2.05
Weighted Sum with
Fast-pass Only
Average IM240 Test
Time with Fast-pass
Only-
passing
16
2
1
0
3
3
65
210
45
102410
207 see
Number Time *
Number of Number
of Vehicles of
Vehicle Fast- Vehicles
Fast- Fast-fail Outpoints s Fast- passing with Fast
>HC/CO/NOx failing and Fast- Result
failing
>3.405/56.72/7.30 15 31 1054
> 1.891/47.30/4.63 22
> 1.648/38.09/3.58
>1.536/41.09/3.19
> 1.518/36.78/3.02
> 1.296/30.34/2.57 11
>1.120/25.22/2.65 11
>0.915/18.06/2.33 35
>0.805/15.05/2.05 33
24
8
9
9
14
76
245
78
1440
592
837
1017
2156
13148
50470
18642
Weighted
Sum
Average
IM240
Test Time
89356
180 sec
These results indicate that the test time for the IM240 can be reduced by
25% when fast-pass/fast-fail criteria are applied and a reduction of over half
a minute occurs when only fast-pass criteria are applied. Using only fast-
pass criteria allows for the collection of diagnostic data so that failing
cars may be repaired more effectively.
Because Hammond cars with second-by-second data were typically shut off
for 10 minutes, catalyst cool down could have caused high emissions during the
early parts of the test and adversely affected fast-pass and fast-fail.
Similarly, vehicles that drive a short distance to an I/M station may not be
fully warmed up when they start the test. Therefore, additional analyses were
L-5
-------
performed without integrating over the first part of the IM240. In effect,
utilizing the first segment of the IM240 as preconditioning. Three different
integration starting points were used. Since the accelerations contribute the
most toward catalyst light-off, these starting points follow the first three
accelerations of the IM240 cycle. The integrations begin after 17, 35 and 47
seconds of the test. The results of these analyses are displayed here.
Number
Fast-pass Outpoints of
Number Time*
Number of Number of
of Vehicles Vehicles
ode*
1
2
3
4
5
6
7
8
9
Tune
> (sec)
17-34
17-60
17-74
17-93
17-113
17-154
17-173
17-206
17-239
Purge
<0.525/0.95/1.33
^/\ t
•^ v» A
<0.504/0.54/1.10
>0.3
<0.465/0.90/0.96
>0.3
<0.400/0.90/0.88
>0.4
<0.486/0.91/0.93
>0.5
<0.593/1.09/1.00
>0.6
<0.641/3.08/1.38
>0.7
<0.826/15.33/1.82
>0.8
£0.805/15.05/2.05
Vehicles
Fast-
passing
11
1
4
0
5
3
56
217
48
Fast-fail Outpoints
>HC/CO/NOx
>2.643/76.94/10.33
> 1.892/53. 86/5. 11
>1. 615/41. 40/3.77
> 1.498/45.64/3.27
>1. 484/40. 16/3.08
>1.265/32.27/2.66
>1.080/26.48/2.71
>0.936/18.44/2.32
>0.805/15.05/2.05
Vehicles
Fast-
failing
19
11
11
10
7
10
10
37
34
Fast-
passing
and Fast-
failing
30
12
15
10
12
13
66
254
82
with Fast
Result
1020
720
1110
930
1356
2002
11418
52324
19598
Weighted Sum with
Fast-pass Only 103230
Average IM240
Test Time with
Fast-pass Only 209 sec
Weighted
Sum 90478
Avenge
IM240
Test Time 183 sec
L-6
-------
Number
Fast-pass Outpoints of
Tune Purge Fast-
passing
Number
of
Vehicles
Fast-fail Outpoints Fast-
>HC/CO/NOx failing
Number Time *
of Number of
Vehicles Vehicles
Fast- with Fast
passing Result
and Fast-
failing
1
2
3
4
5
6
7
8
9
N/A
35-60
35-74
35-93
35-113
35-154
35-173
35-206
35-239
N/A
<0.493/0.79/0.90
>0.3
<0.403/0.73/0.79
>0.3
<0.340/0.69/0.75
>0.4
<0.454/0.9 1/0.82
>0.5
<0.585/1.10/0.93
>0.6
<0.575/2.85/1.37
>0.7
<0.795/15.17/1.84
>0.8
SO.805/15.05/2.05
£1.0
Weighted Sum with
Fast-pass Only
Average IM240
Test Time with
Fast-pass Only
N/A
19
4
2
5
2
48
221
44
102452
207 sec
N/A
>1.983/41.71/3.71
>1.499/31.32/3.08
> 1. 450/55.7 1/3.09
>1.406/47.21/3.07
>1.299/35.99/2.59
>1.061/28.83/2.81
>0.966/19.48/2.37
>0.805/15.05/2.05
N/A
41
8
5
3
7
7
35
43
N/A
60
12
7
8
9
55
256
87
Weighted
Sum
Average
IM240
Test Tune
N/A
3600
888
651
904
1386
9515
52736
20793
90473
183 sec
L-7
-------
Fast-pass Outpoints
Tune Purge
1
2
N/A
47-60
3
4
5
6
7
8
9
47-74
47-93
47-113
47-154
47-173
47-206
47-239
N/A
<0.458/0.40/1.05
>0.3
<0.375J0.46/0.83
>0.3
<0.310/0.52/0.76
>0.4
<0.434/0.94/0.85
>0.5
<0.594/1.14/0.96
>0.6
<0.550/2.88/1.43
>0.7
<0.751/14.82/1.91
>0.8
SO.805/15.05/2.05
£1.0
Weighted Sum with
Fast-pass Only
Average IM240
Test Tune with
Fast-pass Only
Number Time *
Number Number of Number of
of of Vehicles Vehicles
Vehicles Vehicles Fast- with Fast
Fast- Fast-fail Outpoints Fast- passing Result
passing >HC/CO/NOx failing and Fast-
failing
N/A N/A N/A N/A N/A
6 >2.089/37.20/3.67 41 47 2820
102119
207 sec
>1.282/33.21/2.99 14
>l.737/67.22/3.10
15 > 1.619/54.68/3.08
>1.355/39.55/2.55 8
48 >1.095/30.98/2.82
220 >1.004/20.39/2.42 35
38 >0.805/15.05/2.05 45
23
1702
465
17
12
52
255
1921
1848
8996
52530
83
19837
Weighted
Sum 90119
Average
IM240
Test Time 182 sec
These results indicate, that for the data used in this analysis,
preconditioning has little effect on the average test time of the fast-
pass/fast-fail algorithm used. In spite of this, these estimates are
considered conservative for several reasons. First, older cars are excluded
from the analysis. Since most grossly emitting vehicles are older vehicles,
the inclusion of these cars would be expected to increase the number of fast-
failing vehicles and reduce the test time further. However, this reduction
may be offset by a reduction in the percentage of vehicles fast-passing. More
important than the vehicle sample is the algorithm used. If a continuous
function were used, actual test times could be used to calculate the average.
This should lead to significant time savings compared to using the last second
of a particular mode as the required test time for all vehicles that pass or
L-8
-------
fail during that mode, it is unlikely that all the vehicles failing or
passing a particular mode would have required the full mode to determine their
outcome. Therefore, average test times for vehicles passing the IM240 at
second 60 would be significantly less than 60 seconds. Likewise, this would
be true for each mode. On-going analyses are being performed to investigate
this and other alternatives such as the IM240-reversed. Finally, EPA will
continue to develop alternative algorithms which are also expected to reduce
the average test time for the IM240.
L-9
-------