EPA-HONDA Emission Laboratory
Correlation Study
Environmental Protection Agency
Office of Air and Waste Management
Office of Mobile Source Air Pollution Control
Emission Control Technology Division
Standards Development and Support Branch
Ann Arbor, Michigan
October, 1975
-------
Abstract
A laboratory correlation program between the EPA Motor Vehicle
Emission Laboratory (MVEL) and the new Ann Arbor Honda emission test
laboratory has been completed. Statistical analysis of data from a
1976 Honda Civic CVCC tested for exhaust emissions and fuel consump-
tion at two inertia settings indicate no differences in measured values
of CO, C02» and calculated fuel consumption. A statistical difference
exists between measured quantities of HC and NOX.
-------
1. Introduction
American Honda Motor Company, Inc. has recently completed con-
struction and the facilities checkout of their light duty vehicle emis-
sions test laboratory located at 3947 Research Park Drive, Ann Arbor.
After one test cell became operational about mid-August 1975, American
Honda requested that a laboratory to laboratory correlation program be
conducted to locate and correct potential significant differences be-
tween test results from EPA and Honda.
On Friday, September 12, a vehicle was tested at the Honda Ann
Arbor laboratory, and on Monday September 15, the same vehicle and test
sequence were used to gather emission and fuel economy data at the EPA-
MVEL. The data gathered at these two test sites were then used for
comparison purposes.
2. Test Design
2.1 Test Sequence
The laboratory correlation program consisted of hot start exhaust
emission tests, EPA highway fuel economy tests, dynamometer coastdown
checks, and static gas checks of HC, CO, C0_, and NOx.
-------
-2-
The test sequence, as repeated at each site, consisted of the following
tests and checks:
1) @ 2000// IW 2 Hot-start LA-4 tests (2-bag)
1 HFET
Dynamometer coastdown checks
2) @ 2250// IW 2 Hot-start LA-4 tests
1 HFET
Dynamometer coastdown checks
3) Gas analysis of HC (propane), CO, C0», and NOx dilute concentrations.
A total of 12 cylinders were analyzed at each laboratory.
2.2 Test Vehicle
The test vehicle was a 1976 model year Honda Civic CVCC, l.i litre,
5 speed manual transmission, equipped to measure and record torque and
speed vs. time. The vehicle was tested at 2000 and 2250 pounds inertia.
The fuel type used was Indolene 30. Honda supplied the driver for all
tests at both laboratories.
2.2-1 Preconditioning;
All exhaust emission and fuel economy tests were hot start tests.
The test vehicle was used to warm-up the dynamometer and set the correct
value of indicated horsepower. The preconditioning consisted of 15
-------
3
minutes at 30 mph, 3 minutes at 50 mph, and 1 minute at idle.
2.3 Facilities
2.3-1 Equipment; The vehicle was tested at the Honda
Ann Arbor Laboratory using their light duty test cell (only one is
operational at present) and analytical equipment. The test cell has
temperature and humidity control.
At the EPA-MVEL in Ann Arbor, dynamometer 5, CVS 25C, and analy-
tical train 21 were used for the correlation tests. A complete des-
cription of the Honda and EPA equipment is presented in Table 1.
2.3-2 Calibration; Analyzer, CVS, and dynamometer checks and
calibration were performed in accordance with current Federal Register
requirements. Honda and EPA propane recovery checks and chemiluminescent
converter efficiency checks were within acceptable guidelines.
3. Test Results
3.1 Emission and Fuel Economy Results
Test results are summarized in Table 2. Values for ₯_ (bag 1),
Hi
and Yuc, (bag 2) have units of grams per phase. Composite values are
tlD
computed in grams/mile. NOx results are reported as corrected values,
-------
-4-
Table 1
TEST SITE INSTRUMENTATION
Analyzers
EC
LCO
HCO
co2
NOx
Analyzer Bench
CVS
Dynamometer
Driver's Aid
EPA #21
Beckman 400 .
0-50, 100, 250 ppm
H2/N2 fuel
Bendix 8501
0-50, 500
Beckman 315A
0-1000, 2500
MSA 202
0-2.5, 5%
TECO 10
0-100, 250
Homebuilt
Aeronutronic.
Ford CFV
Clayton CTE-50
Flywheel Driven Inertia
Simulation
Varian, 5" FS = 60 mph
4"/min. Preprinted Trace
Honda
Beckman 400
0-50, 100 ppm
H2/N2 fuel
Beckman 315B
0-200, 400, 800
Beckman 315B
0-1000, 2000, 5000
Beckman 315B
0-2.0, 4, 6%
Beckman 951
0-100, 400
AESi
AESi PDF
Clayton CTE-50
Flywheel Driven
Inertia Simulation
AESi, 5" FS = 60 mph
4"/min. Preprinted
Trace
Computer
IBM 370, Off-line
NOVA/2
-------
Table. 2
Date
Test Site
September 15
EPA-MVEL
Inertia Weight Ibs.
Baro. Press, mm
Hg
Dry Bulb/Wet Bulb
°c
Relative Humidity
Z
NOx Correction
Factor
2 Bag LA-4 Test
Hot Start
YRT (g/phase)
Ys (g/phase)
comp
LA-4 FE (mpg)
HFET (mpg)
HC
CO
co2
NOx
HC
CO
co2
.NOx
HC
CO
.CO,
NOx
2000
747.3
22.2/15.5
50
0.928
2.40
16.58
950.9
7.53
2.96
13.07
1171.8
4.36
0.71
3.95
283.0
1.58
30.5
2000
747.5
22.5/15.5
50
0.928
2.11
13.05
953.1
7.60
2.86
12.95
1195.2
4.32
0.66
3.47
286.4
1.59
30.2
40.0
2250
747.5
22.2/15
50
0.924
2.31
16.28
952.4
7.88
2.95
13.68
1207.2
4.87
0.70
3.99
287.9
1.70
29.9
2250
747.5
.5 22.7/15.8
50
0.930
2.28
17.03
967.0
8.22
2.95
13.97
1198.5
5.09
0.69
4.13
288.7
1.77
29.8
39.1
2000
737.1
22.8/16.
50
0.966
2.36
12.75
963.7
7.84
3.24
12.72
1181.9
4.75
0.74
3.34
286.1
1.67
30.2
September 12
HONDA-ANN ARBOR
2000
737.1
7 23.9/16.7
51
0.959
2.23
13.15
945.2
8.41
3.24
12.85
1173.0
4.92
; 0.72
3.46
282.4
1.77
30.6
39.2
2250
737.1
22.8/16.1
50
0.946
2.72
18.36
976.9
9.16
3.31
13.94
1208.1
5.94
0.80
4.30
291.3
2.00
29.5
2250
. 737.1
22.2/16.1
53
0.952
2.72
17.77
980.0
8.81 .
3.18
13.55
1177.9
5.55
0.78
4.17
287.7
1.91
29.9
38.5
j
Ul
-------
Composite values of CO and C0« were repeatable and showed accept-
able correlation at both inertia settings at the two test sites but
Honda measured higher quantities of HC and NOx for all four emission
tests. The calculated fuel consumption for both LA-4 and HFET cycles
agreed well at both test sites.
3.2 Gas Cylinder Analysis
Twelve gas cylinders were analyzed on the same analytical systems
that were used for emission testing. The analyses of the twelve concen-
trations of gases are presented as Table 3.
Only NOx analyses showed substantial deviation between Honda and
EPA. The differences between determinations of NOx concentrations are
also apparent when examining emission results.
3.3 Coastdown Checks
Dynamometer coastdowns were run at each inertia setting at both
test sites. Differences in frictional horsepower losses were not signi-
ficant. Torque requirements at six speed conditions were well within
the range of torque variation observed during the 1975 EPA-MVMA Correla-
tion Study.
-------
-7-
Table 3
GAS ANALYSIS
Cylinder 'Concentration (ppm) H-E
Nominal . EPA Honda ~T" x WO/°
HC 25.1 . 24.3 24.9 +2.46
73.2 72.9 74.0 +1.50
CO 86.0 82.2 81.0 -1.45
450.0 451.3 440.0 -2.50
C02 (%) 0.65 0.664 0.645 -2.86
1.31 1.287 1.300 +1.01
NOx 17.1 15.4 16.1 +4.54
46.6 43.4 ' 47.3 +8.98
88.6 88.3 95.9 +8.60
13.9 11.2 11.8 +5.35
41.7 36.6 39.7 +8.46
83.7 75.8 82.2 +8.44
-------
-8-
4. Statistical Treatment
4.1 Emission and Fuel Economy Results
The mean values for HC, CO, CO-, NOx and fuel economy results were
examined to check for significant statistical differences between Honda
and EPA. An independent "t" statistic was calculated for each pair of
values at each inertia weight. At 2000 pounds inertia a .statistically
significant difference was observed at the 90% confidence level for
measured values of HC and NOx, with Honda measuring 8% higher values of
HC and NOx. At 2250 pounds inertia Honda measured 13% higher values of
HC and NOx. Again, these differences were only significant at the 90%
confidence level. No significant differences were apparent when exam-
ining CO, C02, and LA-4 and HFET fuel economy. Statistical analyses are
summarized in Tables 4 and 5.
4.2 Discussion of Test Results
An examination of the EPA HC and CO results reveals a higher degree
of variation among results at 2000 pounds inertia than at 2250 pounds
inertia. There is no inherent reason for higher variability at the
lower inertia setting. The high variability among EPA results is due to
the high HC and CO values measured for test 1 at 2000 pounds inertia which
can be traced to the bag 1 concentrations. At this time the high values
for HC and CO cannot be accounted for and must be attributed to test-
to-test variability. The bag 1 HC and CO data were closely examined
-------
-9-
Table 4
STATISTICAL ANALYSIS
2000// IW
EPA 1)
2)
x
a
a, % of x
HONDA 1)
2)
x
a
a, % of x
(x - x )
i nn?
J.UU/0
t-statistic
t-value
99% C.L.
6.965
95% C.L.
2.920
90% C.L.
1,886
HC
0.71
0.66
0.685
, , 0.035
' 5.1
0.75
0.73
0.740
0.014
,. 1.9
+8.0
+2.04
ND**
'ND
D*
CO
3.95
3.47
3.710
.. 0.339
9.1
3.34
3.46
3.405
0,092
2.7
-8.2
-1.25
ND
ND
ND
cp_2
283.0
286.0
284.70
2.40
0.84
286.1
282.4
284.25
2.62
0.92
-0.15
-0.18
ND
ND
ND
NOx
1.58
1.59
1.585
0.007
0.44
1.67
1.77
1.720
0.071...
4.13 ,
+8.5
+2.69
ND
ND
D
LA-4 FET
MFC MPG
30.48
30.17
40.05
30.3-25
0.219
0.72
30.21
30.59
39.18
30.363
0,205.
0.68 .
+0.12 -2.3
+0.31
ND
. ND
ND
*Difference exists
**No Difference exists
-------
-10-
Table 5
STATISTICAL ANALYSIS
2250// IW
EPA 1)
2)
X
o
.a, % of S
HONDA 1)
2)
X
a
a, % of x
(Xii - Xp)
h i nnc/
_ J.UUA
XE
t-statistic
t-value
99% C.L.
6.965
95% C.L.
2.920
90% C.L.
1.88fi
HC
.70
.69
0.695
0.007
1.0
0.80
0.78
0.790
0.014
1.8
,+13.7
i
+8.50
D*
D
D
CO
3.99
4.13
4.060
0.099
2.4
4.30
4.17
4.235
0.092
2.17
+4.3
+1.83
ND**
ND
ND
co2
287.9
288.7
288.30
0.57
0.19
291.3
287.7
289.5
2.546
0.87
+0.41
+0.65
ND
ND
ND
NOx
1.70
'. 1.77
1.735
0.049
2.82
2.00
1.91
1.96
0.064
3.26
+13.0
+3.86
ND
D
D
LA-4 - FET
MPG MPG
29.91
29.82
39.13
29.87
0.064
0.21
29.52
29.91
38.49
29.78
0.222
0.74
-0.30 -1.6
-0.75
ND
ND
ND
* Difference exists
** No Difference exists
-------
-11-
for transcription errors, span and zero set point errors, analyzer
or CVS problems, and other possible errors, but none were located.
Possible sources of test-to-test variability such as vehicle, driver,
and ambient condition changes have likewise been examined and ruled out
as the source of variability which would account for the differences in
EPA bag 1 results for HC and CO.
V *
It should be noted that the high degree of EPA CO variability
(traceable to bag 1, test 1 results at 2000 pounds) did not result in a
proven statistical difference between EPA and Honda, even at the 99%
confidence level. The power of the "t" test is related to the accuracy
of the estimated values of standard deviation for each sample. Higher
values of sample standard deviation will result in lower calculated
values of the "t" statistic and consequently a lesser chance of observing
a significant difference between samples. A third test at each inertia
setting is necessary if an accurate estimate of the standard deviation
cannot be attained with two tests.
Differences in NOx can likely be attributed to the NOx analyzer or
NOx calibration gases. Honda and EPA use different NOx analyzers,
although both chemiluminescent analyzers are negative pressure reaction
chamber types, and both analyzers have N0_ -> NO conversion efficiencies
above 96%. The NOx calibration and span gases should be rechecked at
each site. The discrepancies between NOx emission results are not
likely to be due to dynamometer loading because C0_ differences were
negligible and dynamometer coastdowns and torque readings did not indi-
-------
-12-
cate differences in dynamometer power absorption characteristics.
The reasons for hydrpcarbon differences are difficult to assess.
They cannot readily be attributed to ambient effects such as tempera-
ture, humidity, and barometric pressure. Dry bulb temperatures and
absolute humidity were nearly identical for both test sites but baro-
metric pressure was 0.4 in. Hg higher during the tests at EPA. The 1975
EPA-MVMA Correlation Study estimates the effects of changes in ambient
test conditions in terms of changes in grams/mile for HC, CO, CO-, NOx
and fuel consumption. An 0.4 in. Hg change in barometric pressure may
cause a change of 0.066 gm/mile of HC. However, it appears that this
difference should not be entirely attributed to barometric pressure
change because the effect of barometric pressure on CO and CO- is usually
more pronounced than changes in HC levels, but neither CO or CO- show
differences between test sites. No significant differences were observed
between CO, CO-, and fuel consumption measurements at the 99% confidence
level.
5. Conclusions
1) CO, C02, and fuel economy results agree closely between EPA-
MVEL and the Honda-Ann Arbor test laboratory.
2) HC and NOx emission measurements show disagreement between .test
sites. Honda measures 8% higher levels of HC and NOx at 2000 pounds IW
and 13% higher levels of HC and NOx at 2250 pounds IW.
-------
-13-
3) Differences in NOx readings are likely due to analyzer setup
and/or errors in using calibration and span gases.
6. Recommendations
1) NOx analyzer operating parameters should be checked. Span
gases should also be checked for errors in naming concentrations.
- »/ ,
2) Testing at each site on more than one day would probably
result in a greater change in ambient conditions. This would permit
more accurate estimates of ambient correction factors to be made.
3) Honda should include this laboratory as part of the next JAMA
correlation study.
References .
1. Leiferman, M., and Wilson, G.; "1975 EPA-US MVMA Emission Lab-
oratory Correlation Study", Environmental Protection Agency, Office
of Air and Waste Management, Office of Mobile Source Air Pollution
Control, Emission Control Technology Division.
------- |