EPA 600/3-83-104
PB84-120369
Effect of Load Simulation on
Auto Emissions and Model Performance
(U.S.) Environmental Sciences Research Lab.
Research Triangle Park, NC
Nov 83
U.S. Department of Commerce
Katkmai Technical Information Service
-------
TECHNICAL REPORT DATA
(Please read Instructions on :he reverse before completing)
1. REPORT NO.
EPA-600/3-83-104
3. RE
4. TITLE ANP SUBTITLE
EFFECT OF LOAD SIMULATION ON AUTO EMISSIONS AND
MODEL PERFORMANCE
5. REPORT DATE
November 1983
6. PERFORMING ORGANIZATION CODE
7. AUTHORIS)
P. Gabele and R. Snow
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
10. PROGRAM ELEMENT NO.
C9YA1C/04-3080 (FY-83)
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
Environmental Sciences Research Laboratory - RTP, NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NO "ES
the overall objective of this study was to identify sources which might contri-
bute to errors in mobile source emission rate model predictions. The effect of road
load simulation on exhaust emissions was examined and an evaluation of the U.S.
Environmental Protection Agency's Automobile Exhaust Emission Modal Model was
conducted. The Modal Model is a component of the Intersection Midblock Model and
MOBILE2, two widely used programs for predicting emissions from mobile sources.
Results from tests on a Chevrolet Celebrity (3000 pounds gross vehicle weight)
indicated that emissions during tests with water brake load simulation did not
differ significantly from those during tests with actual road load simulation. For
the Celebrity, the load applied by the water brake with the tire rolling resistance
losses on the dynamometer was approximately equal to the actual road load measured
in highway tests.
Evaluation of the Modal Model was completed by comparing actual emissions data
with predicted values. The Celebrity was used to generate emissions data for the
New York City Cycle, the Surveillance Driving Schedule, and the Federal Test Proce-
dure. Results indicated that the Modal Model was unable to accurately predict
emission rates for the Celebrity.
"~
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS ITMi Repartl
UNCLASSIFIED
21. NO. OF PAGES
29
20. SECURITY CLASS
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (R«v. 4-77) PREVIOUS EDITION is OBSOLETE
-------
EPA-600/3-83-104
November 1983
EFFECT OF LOAD SIMULATION ON AUTO EMISSIONS
AND MODEL PERFORMANCE
by
Peter Gabele and Richard Snow
Emissions Measurement and Characterization Division
Environmental Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
BEST
AVAILABLE
COPY
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U. S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
NOTICE
This document has been reviewed in accordarce with
U.S. Environmental Protection Agency policv and
approved for publicaticn. Mention of trade names
or commercial products does not constitute endorse-
ment or recommendation for use.
ii
-------
PREFACE
Ambient, standards for air pollutants exist for the protection of humans
and their environment. Adequate planning is necessary to assure compliance
with standard levels. But unless communities are able to predict in advance
the cause and effect relationships which take place when emission sources
arrive on scene, planning for a clean and healthy environment becomes
unmanageable.
Models for predicting emissions from automobiles have been developed and
promulgated by the EPA. These model:; are used extensively by states to
develop scenarios for environmental planning. The Environmental Sciences
Research Laboratory contributes to the formulation of predictive models by
providing emission factors from mobile sources. Because emission control
technology progresses with time, emission factors and the methodology used to
obtain them must be updated periodically.
This report evaluates the effectiveness in predicting current vehicles'
emissions of an emission rate model which is used widely throughout the United
States. It also examines the dynamometer test procedure which is used to
obtain automobile emission factors.
111
-------
ABSTRACT
The overall objective of this study w
-------
CONTENTS
Preface iii
Abstract iv
Figures vi
Tables vii
Abbreviations viii
1. Introduction 1
2. Conclusions 3
3. Recommendations 4
4. Experimental Procedures 5
Facility description 5
Test vehicle 5
Experimental design 6
Dynamometer road load simulation 6
Dynamometer inertia simulation 7
Emissions testing 8
Test cycle descriptions 8
Computer model predictions 9
5. Results and Discussion 12
Actual road load determination 12
Dynamometer road load simulation 12
Dynamometer water brake simulation 14
Vehicle inertia load simulation 14
Emissions testing 14
Model predictions 15
References 21
-------
FIGURES
Number Page
1 Sampling scheme 18
2 Actual road load curves , 18
3 Actual road load curve, .ictual road load
simulation plus rolling resistance losses on the
dynamorneter, rolling resistance losses on the
dynamometer 19
4 Actual road load and simulated actual road load
curves 19
5 Actual road load versus water brake load
simulation 20
6 Emission rate function curves for accelerations
held constant at 1.5 to 2.25 mph/s 20
-------
TABLES
Number Page
1 Test Vehicle Description 6
2 Test Matrix 8
3 Test Cycle Description 9
4 Means, Standard Deviations, and Estimated
Errors for Emission Rates 11
5 Torque Data Collected During Road Tests 13
6 Vehicle Inertia Measurements for 0 to 60 mph
WOT Accelerations 14
7 Exhaust Emissions Summary 15
8 Percentage Error of Predicted Values 16
9 Percentage Error of Predicted Values 17
-------
LIST OF ABBREVIATIONS AND SYMBOLS
ABBREVIATIONS
A/D analog to digital
CFM cubic feet per minute
CVS constant volume sampler
FTP Federal Test Procedure
HC(s) Hydrocarbon(s)
I MM Intersection Midblock Model
NO nitrogen oxides
NYCC New York City Cycle
SDS Surveillance Driving Schedule
SS steady state
rpm revolutions per minute
VMT vehicle miles traveled
-------
SECTION 1
INTRODUCTION
>
Ambient air concentrations of carbon monoxide (CO) within urban areas are
often significantly higher than those predicted by dispersion models. Because
CO is emitted predominantly by motor vehicles, models used to predict CO
emission rates from groups or classes of motor vehicles could be contributing
substantially to the shortfall ard should be examined for obvious
inaccuracies. Two widely used models or r,-ograms for estimating emissions
concentrations for hydrocarbons(HCs), nitrogen oxides (NO ), and CO are
MOBILE2 (1) and the Intersection Midblock Model(IMM) (2). x
MOG1LE2 was developed by EPA using data acquired over 12 years of
emissions testing. Because the bulk of emissions data was collected for the
Federal Test Procedure (FTP), other models are used within MOBILE2 to correct
emission rates to non-FTP conditions. For example, the EPA Modal Analysis
Model is used to estimate emission rates for vehicles operating over driving
cycles with average speeds different from the FTP. These emission rates are
used to develop speed correction factors for correcting FTP data to the
specific case being modeled.
The Intersection Midblock Model ([MM) was also developed by EPA to aid in
the identification and analysis of CO hot spot locations. It uses the EPA
Modal Analysis Model to calculate CO emissions due to vehicle cruising,
acceleration-deceleration (accel-decel), and assigns these emissions to
traffic links based upon calculated intersection parameters. After emissions
have been distributed among individual lanes of each link, the EPA HIWAY Model
is used to predict CO ambient concentrations at the desired locations.
One questionable component of both models is the EPA Modal Analysis Model
which is designed to predict emission rates for specific vehicles being
operated over any defined driving schedule. The Modal Analysis Model was
developed in 1973 using data obtained on pre catalyst cars. Altnough it was
later refined and updated to 1977 model-year cars, the model remains outdated
in the context of automotive pollution control advances which have occurred
since that time.
Additionally, the emissions used to develop the Modal Model were obtained
from tests on water brake rather than electric dynamometers. Although the
test procedures using water brake dynamometers is an adequate method for
emissions certification, it is unabla to simulate vehicle road loads as
accurately as an electric dynamometer (3,4).
Both the Modal Model and dynamometer load simulation technique should be
examined for obvious inaccuracies. If inaccuracies exist, compensatory
methods or techniques can, hopefully, be applied to reduce errors and improve
model quality.
The objectives of this experimental program are to investigate
contentions that water brake dynamometers fail to accurately simulate vehicle
road loads during tests and to evaluate the Modal Model's ability to
effectively predict emission rates for new cars. The effect of water brake
-------
dynamometer loading on exhaust emissions was measured for three driving
cycles: the FTP, the Surveillance Driving Schedule (SDS), and the New York
City Cycle (NYCC). For each cycle, emission rates measured at actual road
load were compared with those measured for water brake load. Examination of
the Modal Model involved comparing emission rates predicted by the model with
those actually measured. Predictions made for Test Phases 2 and 3 of the FTP,
the SDS, and NYCC were compared with measured emissions data that had been
obtained during the load study phase of the program.
-------
SECTION 2
CONCLUSIONS
Investigation of the effect of dynamometer load characteristics upon
regulated emission rates and an evaluation of the EPA Modal Analysis Model
were completed. Based upon the study's findings the following is concluded:
1. The load applied by the water brake and the tire rolling resistance
losses on the dynamometer was approximately equal to the actual road load
measured in highway tests.
2. Regulated emission rates for the Celebrity are not significantly
different, when tested using the wati;r brake simulation versus actual road
load simulation on an electric dynamometer.
3. The EPA Exhaust Emissions Moda'i Model is an inaccurate predictor of
regulated emissions from the Celebrity.
Ihe conclusions suggest that water orake dynamometers adequately simulate
actual road loads for emissions test purposes. This should hold true for
vehicles such as the Celebrity which have large inertia load components
relative to aerodynamic load components. When the aerodynamic load component
becomes a significant portion of the total road load, dynamometer absorbed
power theoretically deviates with speed from the actual road load. Tne
tendency for this occurrence, which makes simulation of road loads with water
brake dynamometers more difficult, increases for extremely lightweight cars.
Because most data collected for use in MOBILE2 have been from vehicles
roughly equal in size to or larger than the Celebrity, inaccuracies in load
simulation do not have any significant effect on the accuracy of MOBILE2.
However, should minicars (<2000 Ib) ever occupy a significant percentage of
the vehicle miles traveled (VMT), a re-evaluation of dynamometer load
simulation will become necessary.
With regard to the Modal Model evaluation, results in tests on only one
vehicle cannot in themselves disprove the model. This is true because the
model was recommended for prediction of vehicle group emissions and not
individual vehicle emissions (5). However, because high tech emission control
systems have changed the relationship between vehicle speed and emissions
since the model's development, the Modal Model should be updated.
-------
SECTION 3
RECOMMENDATIONS
It is suspected that the Modal Model is an inaccurate predictor of
regulated emissions f®)m late model cars which are equipped with high tech
emission control systems. Because this model plays an active role in both
MOBILE2 and the IMM, it should be updated or, if necessary, replaced with an
acceptable alternative method.
In the case of MOBILE2, speed correction factors are now being obtained
using actual emission test results rather than results predicted using the
Modal Model. This requires testing a rather large cross section of vehicles
over test cycles having different average speeds. These data will be used in
developing realistic speed correction factors for use in MOBiLE3, an upcoming
revision of the current MOBILE2.
-------
SECTION 4
EXPERIMENTAL PROCEDURES
FACILITY DESCRIPTION
All emissions tests were conducted with a chassis dynamometer for vehicle
road load simulation. The Horiba model CD6-800/DMA-915 dynamometer simulated
road load by means of a DC electric motor-generator directly coupled to the
front rolls of the dynamometer. A control system was used to vary armature
current to achieve the desired motor torque. In addition to vehicle road load
simulation, the electric drive was capable of simulating vehicle inertia as
1-lb increments. Rolls of the dynamometer are 22 cm (8.65 in.) in diameter
and are coupled during the automatic calibration mode.
Exhaust gases from the test vehicle were directed via a flexible 7.6 cm
(3-in.) stainless steel line to ;i 45.7-cm (18-in.) diameter dilution, tunnel
(Figure 1). A Constant Volume Sampling System (CVS) located at the rear of
the dilution tunnel drew diluent a.nd exhaust gas at a rate of about 700 CFM.
Regulated gaseous emissions and carbon dioxide (C0~) were determined
using standard bag sampling and analysis procedures in accordance with the
Federal Register (6). In selected test runs, these same emission:, were
measured using a real-time computer system to obtain modal emissions data.
The real-time system, which has been previously described (7), centered around
operation of a Texas Instruments 960B minicomputer. Analyzer response times,
which vary with exhaust gas flow rate, were determined with the aid of a flow
measuring device at the engine air inlet. Analog outputs from the gas
analyzers were directed to the computer through analog to digital (A/D)
converters. In addition to gas data, modal calculations of CVS flow rates
corrected to standard atmospheric conditions were also determined.
TEST VEHICLE
The test vehicle used in this study, a 1982 model year, Chevrolet
Celebrity with a 2.5-1, in-line, 4-cylinder engine is described in Tajle 1.
The engine was fitted with throttle-body fuel injection, and engine exhaust
gases were treated in a three-way single bed catalytic converter. The vehicle
was equipped with cruise control, which was used during steady speed testing.
In order to measure torques required to operate the vehicle during road
and dynamometer testing, wheel torque sensors were instrumented on both front
drive wheels. Signals from each Censor were transmitted to a strain gauge
conditioner which provided an analog output signal as well as a calibration
feature. Torque signals were stored on tape using a four channel, frequency-
modulated instrumentation recorder which was powered off the vehicle's DC
system.
-------
TABLE 1. TEST VEHICLE DESCRIPTION
Vehicle Chevrolet Celebrity
Model year 1982
Engine; type L4
Displexement (1) 2.5
Carburetion EFI (Throttle-body)
Emission control 3-Way catalyst
Inertia Weight, (Ibs) 3500
Compressions ratio 8.2
Net HP @ RPM 112/4800
Transmission Auto
Number of Doors Four
Odometer (mi) 10,000
During the road tests a fifth wheel was used to measure vehicle speed.
The speed signal was transmitted inl.o a separate channel on the tape recorder
to enable calculation of load or power since power is a function of torque and
speed.
All testing was done using the Goodyear Viva II steel belted radials with
which the car came equipped. Tire inflation pressures were held at about 35
psi during road tests and 45 psi on the dynamometer. All tires had been
driven about 10,000 mi before the test program began.
EXPERIMENTAL DESIGN
The experimental program involved three stages: (1) electric dynamometer
simulation of actual road and water brake dynamometer leads, (2) emissions
testing, and (3) computer model predictions. Each stage required completion
before thf following stage could proceed.
Dynamometer Road Load Simulations
In the initial stage, the actual road load for the test vehicle was
determined through road testing. Wheel torques and vehicle speeds were
recorded in both directions on a level stretch of highway located on U.S.
Route 64 at the Lake Jordan Dam Project. The test section extended for about
2000 ft with a .02 % grade. On the day of testing the wind speeds, which were
measured by hand-held anemometer, at no time exceeded 1 knot. Low wind speeds
and dry weather created nearly ideal conditions for testing road loads.
In order to develop the required speed-load relationship, the loads on
the vehicle were measured using cruise control at steady speeds ranging from
70 to 30 mph in 10-mph increments. An additional load point at 15 mph was
taken without cruise control. All tests were run in both directions and some
of the tests were repeated. Points at 70, 30, and 15 mph were rerun after the
entire test sequence in order to determine test repeatability.
-------
For each of the speeds examined, load (horsepower) values were calculated
as a function of the average torque multiplied by the wheel revolutions per
minute (rpm). Speed-load points were fed into a computer which provided a
best fit quadradic equation using a nonlinear least squares method. The
resultant speed-load equation or curve represented the actual road load case.
A previously described procedure for determining road load with wheel torque
meters (4) was generally adhered to in this study.
After the actual road load equation had been determined, it was
programmed into the electric dynamometer and the vehicle was tested. Wheel
torque and rpm measurements were made so that a new speed-load equation could
be obtained. This represented the actual road load simulated by the
dynamometer plus tire rolling resistance losses on the dynamometer rolls. To
compensate for added rolling resistance losses, a new equation was obtained by
subtracting the tire losses (power measured at the wheels minus power being
absorbed by the dynamometer) from the actual road load equation. The
resulting equation was then used to simulate loads which when added to tire
rolling resistance losses on the dynamometer closely approximated actual road
loads. Some slight adjustments were made to the coefficients of the
aerodynamic and rolling resistance terms to more closely simulate the actual
case.
Once the actual road load curve could be accurately reproduced on the
dynamometer, an equation for programming water brake loads was sought.
Coast-down data from a Clayton water brake dynamometer located at EPA, Ann
Arbor, was used to develop the necessary water brake speed-load relationship.
Horiba, Inc., the electric dynamometer manufacturer, had also provided an
equation which could be used to simulate water brake loads. This curve, while
found to be almost identical to the one derived from data supplied by EPA, Ann
Arbor, was not used in the program except as a verification device.
Dynamometer Inertia Simulation
Most chassis dynamometers employ flywheels to simulate vehicle inertia
loads. More recently, however, electric dynamometers have dispensed with
flywheels and instead use electric simulation. Many versions provide inertia
selections in 1-lb increments, a feature not practical with flywheels. But
the principal argument in favor of flywheel elimination is the obvious
space-saving advantage.
The inertia setting for the Chevrolet Celebrity was 3000 Ib. While this
value was used in emissions certification, it was somewhat less than the total
effective mass (gravitational plus rotating component) as tested on the road.
The rotating component, estimated from data obtained previously on a similar
car (8), plus the weight of the vehicle, test equipment, and test personnel
was about 3500 Ib.
To measure the accuracy of dynamometer inertia simulation, a group of 0
to 60 mph wide-open throttle accelerations were run on both the level road and
the dynamometer. Integrated torque values measured during the accelerations
were compared in each case to determine accuracy of dynamometer inertia
simulation. Because the actual vehicle weight accelerated on the road was
about 3500 Ib., dynamometer inertia simulation was set at this value.
-------
Emissions Testing
In the second stage of the experimental program, emissions tests were run
on the dynamometer. The test matrix which was followed is shown in Table 2.
Exhaust emission rates were measured for three test cycles and two load
conditions. The test cycles were the FTP, NYCC, and the SDS, and the load
conditions were actual road load and water-brake load. Real-time emission
testing was used with the SDS because modal data from that cycle were required
by the model to predict emission rates for other cycles.
TABLE 2. TEST MATRIX*
Actual road Water brake Real time
Test Cycle simulation simulation system
1st Day
FTP X
NYCC X
NYCC X
SDS X X
SDS X X
SDS X X
SDS X X
2nd Day
FTP X
NYCC X
NYCC X
SDS X • X
SDS X X
SDS X X
SDS X X
* Sequence repeated six times
Test Cycle Uescriptions--
Of the three test cycles or driving sequences examined in this study, the
FTP is most familiar to those in the automotive emissions control field. It
represents a typical urban driving schedule which has been adopted by EPA in
its certification procedure. Total distance of the FTP is 7.5 mi and average
speed is 19.6 mph. The cycle contains three distinct phases—cold transient,
hot stabilized, and hot transient-- and each phase has its characteristic
emissions. A more detailed description of the FTP is given in the Federal
Register.
The NYCC (sometimes referred to as the New York City Driving Cycle)
represents a typical Manhattan driving experience. The cycle is characterized
by low speeds, very high accelerations, frequent stops., and a 40% idle time.
Total distance of the NYCC is 1.2 mi with an average speed of 7.1 mph.
-------
The SDS, unlike the FTP and NYCC, does not represent routine driving but
is designed to measure vehicle emissions over a variety of steady state and
transient driving conditions. To accomplish this the SDS contains 37 distinct
modes: 32 at differing accel/decel rates which originate at different speeds
and 5 at steady state speeds of 0, l!i, 30, 45, and 60 mph. Acceleration and
deceleration rates covered within the driving sequence represent the full
range of rates observed in the CAPE-10 car-chase study (9).
Except for the FTP in which the car was started cold following an over-
night soak period, eacli of the cycles was run following a hot soak period of
10 min. A brief summary of the three driving cycles discussed above is shown
in Table 3.
TABLE 3. TEST CYCLE DESCRIPTIONS
Test Cold Avg. ."Stops Total Duration % Time
Cycle Start Speed i:er Mi Distance (min) Idle
FTP Yes 19.6 '2.40 7.5 22.9 19.0
NYCC No 7.1 9.32 1.2 10.0 35.2
SDS No 33.5 0.82 9.8 17.6 11.7
Computer Model Predictions—
Following the emissions testing stage, data from the SDS were available
for use in the Modal Model to arrive at emission rate predictions. The Modal
Model formulated an instantaneous emission rate function for the vehicle,
which was used to calculate second-by-second emissions over any given speed
versus time driving sequence. Integration of these emission rates resulted in
predicted values for the emissions over the entire test cycle or portions
thereof.
The emission rate function which was developed within the model was based
on assumptions that steady state emission rates are a quadratic function of
speed, that acceleration is a perturbation to the steady state emission rate
function, and that quadratic functions of acceleration represent good approxi-
mations to the perturbation (5). Two mathematical expressions are used to
define the emission rate function, one representing the steady state function
and one representing the non-steady state or transient function. Taken
together, the two functions require specification of 12 coefficients -- 3 for
the steady state function and 9 for the transient functions.
Coefficiont specification is accomplished through processing data obtain-
ed in 37 SDS modes. Data from the 5 steady state modes and from the remaining
32 accel/decel modes are used to define the transient function. Predictions
of instantaneous emissions are carried out by joining the two functions with a
weighting function. The weighting function, which is a function of acceler-
ation, allows for a smooth transition between the steady state and the accel/
decel emission rate functions.
-------
Because modal data, repeatabilities are poor, a sample of at least 25 SDS
tests was obtained. For example, the relative standard deviation for 27 CO
samples obtained in Mode 11 during replicate SDS tests was about 38%. Given
27 replicate results with a standard deviation of 38%, there is a 95% cer-
tainty that the mean emission rate calculated for Mode 11 will be within 15%
of the true value. A complete listing of the means, standard deviations, and
estimated errors for emission rates in each of the 37 SDS modes is shown in
Table 4. It is noted that modal data exceeding two standard deviations were
eliminated before the above statistics were performed.
Using mean emission rate values for eacli mode, the Modal Model was used
to predict emission rates for Test Phases 2 and 3 of the FTP, the NYCC, and
the SDS. Emissions predictions for each of the 38 SDS modes were also made.
Predicted values were compcired with measured values obtained in 12 FTP, 16
NYCC, and 27 SDS tests.
10
-------
TABLE 4. MEANS, STANDARD DEVIATIONS, AND ESTIMATED ERRORS FOR
EMISSION RATES. MEASURED IN 27 REPLICATE SOS TESTS.
Mode
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27*
28
29
30
31
32
33
34
35
36
37
Mean
g/mi
10.5
0.7
8.8
4.3
6.8
0.6
173.2
3.7
33.5
0.6
43.6
0.6
33.0
0.6
0.1
0.9
4.1
0.1
22.6
0.1
137.0
0.8
2.1
14.2
0.7
0.1
6.3
0.6
1.9
73.5
1.1
0.1
1.7
1.3
1.2
1.0
3.7
Std. Dev.
g/mi
5.2
0.6
5.2
2.1
3.7
0.5
31.1
1.4
22.0
0.2
16.6
0.2
17.2
0.3
0.1
0.7
2.7
0.1
15.7
0.1
29.8
0.3
1.2
12.1
0.4
0.2
7.3
0.3
1.6
21.5
0.5
0.1
0.6
0.7
0.5
0.3
1.6
Estimated
error
18
32
22
18
20
33
6
14
24
14
14
13
19
21
43
30
24
34
26
24
8
15
22
32
22
57
44
20
31
11
18
36
13
20
17
11
17
*Estimated error = (Error/mean) x 100%.
11
-------
SECTION 5
RESULTS AND DISCUSSIONS
ACTUAL ROAD LOAD DETERMINATION
Wheel torque data collected during road tests are shown in Table 5. The
numbers represent average torque measured during the test run. Slightly
higher values shown in runs headed north are related to a small (0.02%)
positive grade in that direction. The torque data were collected at steady
speeds ranging from about 15 mph to 70 mph. When the data were reduced, two
speed-load curves were drawn up: on; for speeds < 70 mph and one for
speeds < 50 mph. Figure 2 shows the two curves, which are similar in shape,
alongside each other. Equations for Curves A (<70 mph data) and B (<50 mph),
respectively, are:
T = 37.2 - 0.17 v + 0.024 v2 (A)
T = 26 + 0.33 v + 0.008 v2 (B)
where: v = velocity in mph
T = torque in ft-lb
At 30 mph there is only a 2 ft-lb difference in torque (3.6%) between the two
curves and at 50 mph a 3 ft-lb. difference (3.4%). Because most test cycles
run in this program were at speeds <50 mph, Curve 8 was selected to provide
actual road loads required for simulation.
DYNAMOMETER ROAD LOAD SIMULATION
Curve B was programmed into the dynamometer and the vehicle was readied
for testing. With wheel torque meters in place, torque measurements were
taken over the same steady speed points examined on the road. The speed-load
relationship obtained is shown in Figure 3 as Curve C. The difference between
Curve C and Curve B is due to tire rolling resistance losses on the
dynamometer and is plotted as Curve D in Figure 3. When Curve D was subtracted
from Curve C and the resulting relationship was programmed into the
dynamometer, wheel torque values were again obtained and, after some slight
adjustments to the dynamometer load equations, these values were plotted as
Curve E in Figure 4. This load curve very closely simulates the load curve
(Curve B) obtained in actual road testing.
12
-------
TABLE 5. TORQUE DATA COLLECTED DURING ROAD TESTS
Speed
(mph)
68.9
67.6
70.9
67.9
67.1
59.4
58.9
49.5
49.0
40.3
39.4
31.7
30.1
30.9
28.9
14.4
15.0
14.7
14.3
«,^, — ,»«•-• — -..- — — — ^-.w.^ — — -sa-» — — — —
Direction
N
N
S
S
S
H
S
N
S
N
S
N
N
S
S
N
N
S
S
— — — •» — — .-•. — — .» — ••-.-. — — —
Torque
(ft-lb)
150
143
145
135
128
114
106
88
82
76
67
61
65
53
52
41
41
40
36
13
-------
DYNAMOMETER WATER BRAKE SIMULATION
The dynamometer manufacturer had furnished a load equation which could be
used to simulate a water brake dynamometer. In addition, coast down data were
available from one of the Clayton water brake dynamometers at the EPA in Ann
Arbor. Since both curves were similar, it was arbitrarily decided to use the
relationship developed from the EPA, Ann Arbor coast down data.
Wheel torque measurements obtained while the vehicle was undergoing
water-brake-simulated loads accounted for tire rolling resistance losses on
the dynamometer rolls in addition to dynamometer load. The load curve
representing this condition is shown in figure 5 with the actual road load
curve previously obtained. While: differences do appear at the low and high
speed ends, overall the curves an; not significantly different.
VEHICLE INERTIA LOAD SIMULATION
As a means of insuring accurate dynamometer simulation of vehicle inertia
weight, integrated inertias were measured during wide-open throttle (WOT)
accelerations on both the dynamometer and level road. The results are shown
in Table 6. '
• TABLE 6. VEHICLE INERTIA MEASUREMENTS FOR
0 to 60 MPH WOT ACCELERATIONS
Inertia
Weight
(lb)
Time
(sec)
Integrated
torque
(ft-lb-sec)
Dynamometer 3000 16.4 8522
3000 16.6 8569
3500 19.1 10003
3500 19.0 10043
Road
(6 runs) 3500 * x = 19.4 10909
** s = 2.2% 1.7%
* x = the mean of 6 runs.
** s = relative standard deviation.
At simulated inertia of 3000 lb. the integrated torque values fall about 15%
below those at 3500 lb. The values obtained in dynamometer simulations of the
loaded test car (3500 lb) are within 10% of those obtained on the roadway.
Acceleration times are also shown in Table 6. While a decrease of about 15%
again noted in going from 3500 lb to 3000 lb, a difference of only 2% is
observed between dynamometer and roadway acceleration times. Indications are
that the dynamometer is accurately simulating vehicle inertia load.
EMISSIONS TESTING
Emissions of HC, CO, NO , and tO~ for the Celebrity were determined for
three different tests cycles using actual road load ,and water brake load
14
-------
simulations. Because both load curves were similar, significant differences
in emissions rates were not anticipated. Emissions data presented in Table 7
supported the expectation that no nignificant differences in emissions occur
as a result of load parameter.
TABLE 7. EXHAUST MISSIONS SUMMARY
Emission
HC
CO
N0y
n
co2
=================
Load
simulation
Actual road
Water brake
Actual road
Water brake
Actual road
Water brake
Actual road
Water brake
FTP
(g/mi)
0.24(±.04)
0..23(±.04)
6.33(±.89)
6.83(±.26)
1..27(±.07)
1.25(±.09)
341,2(43)
348,,9(±5)
NYCC
(g/mi)
0.77(±.23)
0.76(±.26)
10.91(tl.41)
11.13(±1.77)
2.06(4.35)
2.11(4.35)
656.8(12.8)
557.7(±10.27)
SOS
(g/mi)
0.18(4.03)
0.17(4.03)
15.84(±1.97)
16.62(42.3)
0.91(4.05)
0.93(4.06)
308.5(±2.6)
307.8(±3.9)
Modal emissions data revealed no significant differences betv/een the two
imposed load conditions. However, large scatter, characteristic of data
obtained in replicate modal analysis runs, completely masks any differences
which might have existed modally (see Table 4). The purpose of obtaining
modal data during the SDS was not so much to examine emissions differences
because of load parameter changes as to evaluate the accuracy of Modal Model
predictions.
MODEL PREDICTIONS
Emission rates for HC, CO, and NO were predicted for Test Phase 2 and 3
of the FTP, the NYCC, the SDS, and eacfh of 33 modes of the SDS. Predictions
were not made for Test Phase 1 of the FTP because cold start emissions cannot
be successfully modeled by the Modal Analysis Model.
Table 8 shows the percentage error of predicted values compared to those
actually measured. The negative values indicate that in all cases predictions
are lower than measured values. For CO there is a trend of increasing error
as average cycle speed increases. Generally, the error in predicting NO
emission rates is lower than those for predicting CO and HC.
CO emission rate predictions for each SDS mode were compared with the
calculated emission rates for those same modes. A summary of the results
showing the error of the prediction is shown in Table 9. The negative values
again indicate that with only one exception predictions are lower than
measured values. For acceleration modes with average speeds over 30 mph
predictions appear to be unreasonably low. Measured emission rates for these
modes were the highest for the entire driving cycle while the predicted values
15
-------
were usually less than those for deceleration modes. Predictions for idle and
15 mph SS were very close to the measured values; however, those for 30 and 45
mph SS were rather low and high, respectively.
In general, predictions made by the: EPA Modal Analysis Model were
considerably lower than measured values. This was particularly apparent with
CO predictions for driving cycles and SDS accel modes having average speeds
over 30 mph. To further investigate this trend, a family of curves (see
Figure 6) describing the emission rate function were drawn showing CO
emissions as a function of velocity for accelerations ranging from 1.5 mph/s
to 2.25 mph/s. For the Celebrity negative emission rates occur at frequent
points within the model. Emissions, regardless of acceleration, appear about
the same at 30 mph. Since these are physically impossible trends for the
vehicle tested, the curves illustrate an obvious discrepancy in the model.
The percent errors as given in Tables 6 and 7 include test to test
variability as well as the error in model predictions. Table 3 shows that in
most cases test variability was rather high. It is also noted that the error
percentages in Tables 6 and 7 do not translate directly to MOBILE2 which
merely uses emission rate prediction to arrive at speed correction factors.
TABLE 8. PERCENTAGE EKROR OF PREDICTED VALUES TO ACTUAL MEASURED VALUES
HC CO NOx
Cycle Avg. Speed % error* % error % error
(mph)
NYCC
FTP Phase 2
FTP Phase 3
SDS
7.1
16.
25.6
33.5
-83
-47
-63
-64
-36
-46
-65
-90
-63
0
-27
-39
* % err/r = '(predicted-measured) r measured] x 100%
16
-------
T/IBLE 9. PERCENTAGE ERROR OR PREDICTED VALUES
Modes CO, % error*
Accel s >30 mph (10 rnocL>s) -53
Accels <30 mph (7 modes) -77
Decels (15 modes) -54
Idle -2
15 mph SS** -2
30 mph SS -70
45 mph bi +50
60 mph SS -15
* % error = ((predicted-measured) f measured) x ".00%.
** SS = Steady speed.
17
-------
Ftgurt 1. Simpltng
figure
2. Actual rojd '>oh *nd below. B for ttO Epn *r,.
-------
160
140
I2B
100
•to
VELOCITY
rtfurt ].
Actual road load curva (I), actual ru«l load llnulltlon plut rolling redltlncc totfttt
On the djMrOMUf (c) > r«IMMQ rtltitarxt loll«« on tfjrnjfiocvttr (C-tl*l>).
i ear
flgwre «.
Iw! tni i-wli:-< ;-.;,.•! r>
-------
100
VELOCITY
Floura S. Actiill rcil lM-J (0) veriui t.*ter bnlt "io*
-------
------- |