Report No. SR91-09-02
Development of an Improved
Computer Simulation of Vehicle
Emissions During Cold Start
and Warm-Up Operation
prepared for:
U.S. Environmental Protection Agency
Office of Mobile Sources
Certification Division
September 30, 1991
prepared by:
Sierra Research, Inc.
1801 J Street
Sacramento, California 95814
(916)444-6666
-------
-------
Development of an
Improved Computer Simulation
of Vehicle Emissions
During Cold Start and Warm-up Operation
Table of Contents
1. Summary 1
2. Introduction 3
Background 3
Work Plan Summary 5
3. Results 7
Data Collection and Analysis 7
VEHSIME Code Development 14
Testing the Algorithm 15
Conclusions/Recommendations 17
Appendix A — A Description of Sierra's Vehicle Emission Simulation Model
-------
Development of an
Improved Computer Simulation
of Vehicle Emissions
During Cold Start and Warm-up Operation
1. Summary
Based on an analysis of modal emissions data from vehicles tested by
EPA, an algorithm has been developed to simulate cold start and warm-up
operation using a vehicle emissions simulation model called VEHSIME.
During the first 450 seconds of operation following a cold start, the
algorithm applies correction factors to the emission rates contained in
"map" of emission rates over the full speed and load range of a warmed-
up engines. The correction factor increases "engine-out" HC and CO
emissions in proportion to the number of seconds left in the warm-up
period. The algorithm also applies a catalyst efficiency factor to the
map of engine-out emissions as soon as the fuel consumption since start
up reaches the cumulative fuel consumption associated with the catalyst
reaching "light-off" temperature on the LA4 cycle (120 seconds into the
cycle). The catalyst efficiency factor is computed from the difference
between the engine-out and tailpipe emission maps for the engine being
used for the simulation.
Figure 1 shows how the Federal Test Procedure (FTP) emissions results
for four vehicles tested by EPA compare to emissions predicted by the
VEHSIME model for two hypothetical vehicles using two different engines
for which warmed up emissions maps were available. (Modal emissions
test results from the same four vehicles tested by EPA were used to
develop the cold start algorithm added to VEHSIME.) As the figure
shows, the VEHSIME model with the cold start algorithm produces
reasonable estimates of emissions during each "bag" of the FTP. The
differences between the results predicted by the VEHSIME model and the
average of the four vehicles tested by EPA are within the range of the
differences between the four vehicles themselves.
Although further data are needed to determine how accurate the new cold
start algorithm is for cycles other than the FTP, the algorithm in its
present form already provides a useful tool for estimating the emissions
during periods of travel from a cold start that are shorter in length
than the first bag of the FTP (3.59 miles). This capability is
necessary for accurately estimating emissions in parking lots and other
places where emissions are likely to be substantially higher than the
average emissions that occur during the first eight minutes of operation
on the LA4 cycle. The ability to accurately predict short trip emission
results can also contribute to improved estimates of emissions
inventories for whole metropolitan areas through the use of more
detailed information regarding the distribution of trips by trip length.
(Under the current procedures for estimating vehicle emissions, vehicle
emissions are assumed to be independent of trip length.)
-1-
-------
Figure 1
Vehicle Test Results
vs. VEHSIME Model
(HC)
EPA 4-Car Sample
VEHSIME 2-Car Sample
EPA 4-Car Sample
VEHSIME 2-Car Sample
Bag 2
Bag 3
-2-
-------
2. Introduction
Under Contract No. 68-C9-0053, Sierra Research, Inc. (Sierra) provides a
variety of analytical services for the Certification Division of the
U.S. Environmental Protection Agency's Office of Mobile Sources. Under
Work Assignment No. 1—02 of the contract, Sierra was directed to analyze
emissions data previously collected by EPA and to use the results of the
analysis to develop an improved technique for simulating vehicle
emissions during cold start and warm-up operation using a vehicle
emissions simulation model called "VEHSIME" (pronounced "vee'-syme").
Background
One concern with the representativeness of the current test procedure
for the emission testing of light-duty vehicles is the extent to which
it adequately simulates cold start emissions. With properly maintained
late-model cars and light trucks, the emissions that occur prior to
catalyst light-off are a substantial portion of the total emissions for
the LA4 cycle. The operation of the vehicle during warm-up obviously
affects how fast the catalyst reaches operating temperature as well as
how much is emitted through the bed of the cold catalyst. The LA4 cycle
subjects the vehicle to relatively high—speed freeway operating
conditions within three minutes of the cold start. Intuitively, this
appears to be a very short period of time between cold start and high
load operation which might unrepresentatively affect catalyst warm-up.
Since vehicle emissions are extremely sensitive to vehicle operating
conditions, obtaining accurate estimates of emissions in various
geographic areas depends on the ability to estimate the effect of
driving pattern differences between regions. Historically, EPA has
attempted to estimate area-specific vehicle emissions through the
application of speed and temperature correction factors to data
generated with the standard test procedures. Such factors are
incorporated within the MOBILE4 computer model. However, the correction
factors built into the model were developed by interpolating and
extrapolating "hot start" test results for the LA4 compared to other
speed-time traces, such as the New York City Cycle and the Highway
cycle. The speed correction factors built into the MOBILE4 model are
based on the assumption that the significance of cold start emissions is
independent of variations in driving pattern. To incorporate some
consideration of how changes in driving patterns might affect cold start
and warm-up emissions, EPA has been interested in developing a computer
model that can simulate such cold start and warm-up operation over any
specified speed-time profile.
Under a previous work assignment, Sierra developed a simplistic
technique to estimate motor vehicle emissions during cold start and
warm-up operation using the engine-map-driven emissions simulation model
VEHSIME. (Appendix A contains a description of the model.) In the
-3-
-------
earlier effort, Sierra used two engine maps, one for tailpipe emissions
of a catalyst-equipped engine and one for non-catalytically treated
("raw") emissions from the same engine, and switched from the non-
catalyst (raw) map to the catalyst map during the first few minutes of
vehicle operation. Additional adjustments included increasing the
emissions of the raw map to simulate the higher HC and CO emissions that
occur during cold start operation. The increase in emission rates for
the raw map at time zero was set such that the area under a straight
line drawn between emission rates at time zero and the warmed-up
emissions rates at 505 seconds would be representative of the typical
ratio of "bag 1" to "bag 3" emissions observed from the testing of non-
catalyst vehicles (i.e., 2.23 for HC and 2.50 for CO). This is
illustrated graphically in Figure 2. In the case of CO, the emissions
rate at time zero for each element of the engine map (represented by
"E", in the figure) was set so that the area enclosed by the polygon
ABDE was 2.50 times the area enclosed by the rectangle ABCD.
Figure 2
Original Cold Start Simulation Method
0>
i
c
o
'w
.52
HI
Catalyst (warm)
Cross-over
(light-off)
505 sec
Time (sec)
-4-
-------
Under the previous work assignment, the VEHSIME model was modified so
that emissions began being computed using the map for catalytically
treated emissions from a warmed up engine at the point necessary to have
the ratio of "Cold FTP/Hot FTP" emissions ([bag 1 + bag 2] + [bag 2 +
bag 3]) equal to the ratio typical for oxidation catalyst equipped cars
(1.89 for HC and 2.25 for CO). That cross-over point turned out to be
122 seconds.
Using the modified version of the VEHSIME model, Sierra investigated the
effect of vehicle driving pattern on emissions during the warm-up phase
of a trip and determined that total vehicle emissions appear to be
strongly influenced by both the driving pattern and the extent to which
tailpipe emissions are elevated during the warm-up phase. Based on
these preliminary results, it became clear that a more rigorous
technique for predicting vehicle emission characteristics during cold
start and warm-up would be desirable for accurately estimating the
emissions from vehicles in customer service under varying driving
patterns.
EPA informed Sierra that increased sophistication of the cold start
simulation added to VEHSIME in the earlier work assignment could
potentially be achieved by utilizing modal emissions data collected by
EPA during 1986 and 1987. Under the EPA testing program, ten different
1984 model vehicles equipped with 3-way catalysts were operated over the
standard "LA4" driving cycle while continuous measurements of emission
concentrations were made upstream and downstream of the catalyst. On
some vehicles, emission measurements were restricted to carbon monoxide.
On other vehicles, both HC and CO were measured. Continuous
measurements of exhaust system temperatures were made at numerous points
in the exhaust system for most of the vehicles. Temperature measurement
locations included upstream of the catalyst, at the face of the
catalyst, mid-bed of the catalyst, and at the exit of the catalyst.
With such data, it is possible to determine how the emissions of- the
engine and the efficiency of the catalyst changed as a function of
elapsed time since start and instantaneous mode of operation. However,
all of the available test data were in strip chart form and not in the
digitized form needed for efficient computer analysis.
Work Plan Summary
The Work Plan developed for Work Assignment 1-02 consisted of three
tasks: 1) Cold Start Simulation Methodology Development, 2) FORTRAN Code
Development; and 3) Reporting.
Under Task 1, one subtask was for the development of the algorithm
concept and the data analysis approach. Another subtask involved the
actual digitizing of strip chart data. The final subtasks were the
analysis of the digitized data and the determination of the most
appropriate algorithms.
Under this Task 2, the algorithms developed under Task 1 were converted
into FORTRAN source code. After debugging and compiling, the code
-5-
-------
changes were tested to determine how well they matched the raw data from
the test vehicles.
The final task involved the preparation of this report and the interim
progress reports submitted each month.
###
-6-
-------
3. Results
Data Collection and Analysis
Second-by-second continuous emissions test data over the LA4 cycle were
acquired from EPA's Catalyst Thermal Activity Test Program (CTAP). Data
for 14 vehicles were received in analog strip chart form. The data
consisted of volumetric emissions and catalyst temperatures (typically
at several locations on the catalyst) as a function of time. Some
testing was conducted at both 75°F and 20°F.
The original expectation was that a comprehensive set of technology— and
emissions—specific adjustment factors could be developed.
Unfortunately, much of the data were incomplete for the purposes of this
analysis because:
• Many of the test packets for the 14 vehicles contained
only catalyst temperature measurements.
• Of the remaining vehicles, only four had both cold and
hot traces for complete LA4—505 cycles with emissions
measured both upstream and downstream of the catalyst.
(Both cold and hot data were required for the cold
start algorithm developed under this effort: a cold
start adjustment to a warmed engine map as a function
of time. Upstream (i.e., engine out) emissions were
needed to develop a cold start adjustment for the
engine separate from the catalyst.)
• The emissions data for the four-vehicle sample
consisted only of CO concentrations. Limited HC
testing was performed on some of the CTAP vehicles but
no HC data were available for the 4 vehicles selected.
No NOx testing was conducted.
Because of the above-mentioned limitations, the dataset suitable for
detailed analysis consisted of the test data described above for the
vehicles shown in Table 1.
Using a SummaSketch II digitizing tablet and Fast Cad software, the data
were digitized from the strip chart traces and loaded onto Sierra's
computer system. For each vehicle in the sample, second-by-second
catalyst efficiencies and cold/hot emission ratios were computed as
follows:
Catalyst Efficiency =
1 - (Tailpipe Emissions -5- Engine Out Emissions)
—7—
-------
Cold/Hot Ratio (C/H) =
Cold Engine Out Emissions / Warm Engine Out Emissions
Table 1
Vehicles For Which Both Hot and Cold Start,
Upstream and Downstream Data Were Available
Vehicle Catalyst
1984 Volvo (2.3L) 3-Way
1984 Volvo (2.3L) 3-Way
1984 Plymouth (2.2L) 3-Way
1984 Chrysler (2.2L) 3-Way
Fuel System
Multi-Pt F.I.
Multi-Pt F.I.
Multi-Pt F.I.
Multi-Pt F.I,
02 Sensor
Heated
Non-Heated
Non-Heated
Non-Heated
An initial review of the computed catalyst efficiencies and C/H ratios
indicated that many of the computed second-by-second catalyst
efficiencies were negative. In addition, initial plots of catalyst
efficiency and C/H ratio showed large amounts of scatter. Further
examination of the emission peaks on the strip charts indicated the
negative efficiencies and high scatter were likely caused by a time lag
between measurement of engine out and tailpipe emissions. Based on the
strip chart data, changes in engine out emissions were being recorded an
average of 6 seconds before changes in tailpipe emissions. Some of this
time delay would be associated with the time it took emissions to travel
from in front of the catalyst to the tailpipe, but most of the delay
could have been associated with differences in the sample trains.
However, the apparent lag varied up to 2 seconds as a function of the
exhaust flow rate, with idle conditions showing the greatest difference
between the time changes in emissions were recorded upstream of the
catalyst and at the tailpipe.
Two separate adjustments were performed on the data in an attempt to
remove the scatter caused by the time lag:
1. A fixed lag of 6 seconds was applied to the tailpipe emission
traces (i.e., tailpipe concentrations were compared to engine-
out concentrations that occurred six seconds earlier) .
2. Emission data for the sample vehicles was re-digitized.
Instead of digitizing emission levels at regular intervals
along the trace (corresponding to the divisions on the time
axis of the strip charts), digitizing was performed at the
midpoint and the endpoint of each of the emission peaks and
valleys as shown in Figure 3. The corresponding points on the
tailpipe traces were then digitized and assigned the same time
as that indicated on the engine out traces.
-8-
-------
Figure 3
"Peak and Valley" Data Digitizing Approach
The efficiencies and C/H ratios were re-computed and plotted for both
adjusted datasets and compared. Both adjustments produced more well-
behaved catalyst efficiencies. C/H ratios computed using the first
approach showed much less scatter than those computed using the second
approach. This was due to the need to interpolate between the digitized
peak/valley points on the hot trace to match the points on the cold
trace. All subsequent analysis of the data was therefore based on the
use of a fixed lag of 6 seconds applied to the tailpipe emission traces.
Figures 4-7 show the catalyst efficiencies for carbon monoxide for each
of the four vehicles. Figures 8-11 show the C/H ratios for the engine-
out emissions from the same cars. Some of the "spikes" of low catalyst
efficiency and high C/H ratio are believed to be associated with the
inaccuracies associated with the assumption that the lag between engine
out and tailpipe emissions measurement was a constant 6 seconds.
The mean catalyst efficiency and C/H ratios computed for the four-
vehicle sample using approach (1) are shown in Figure 12. The data were
further smoothed by computing 10 second "moving averages" at each point.
-9-
-------
Figure 4
1984 Volvo 2.3L, Original Catalyst
Bag #1,75 deg. F FTP Cycle
Temperature and CO Efficiency
0.8
>,0.6
u
c
g>
"o
HI
0.4
0.2
J_
I
1.500
Catalyst
Temperature
CO Efficiency
1.000
CD
0)
500
CD
•
0>
I
Q>
D.
E
0 100
Note: V1N 0156. Packet #13
200 300 400
Time (seconds)
500
600
Figure 5
1984 Volvo 2.3L, Original Catalyst
Bag #1, 75 deg. F FTP Cycle
Temperature and CO Efficiency
O.B
>,0.6
O
c
CD
"o
0.2
1.500
Catalyst
Temperature
CO Efficiency
CT
o>
2
1.000 o)
0)
CD
I
CD
Q.
500 E
0 100
Note: VIN 0192, Packet #17
200 300 400
Time (seconds)
-10-
500
600
-------
Figure 6
1984 Laser 2.2L Turbo, Original Catalyst
Bag #1, 75 deg. F FTP Cycle
Temperature and CO Efficiency
0.8
0)
'o
^
0.4
0.2
.->•
I
1.500
Catalyst
Temperature
CO Efficiency
1.000
a>
0)
TJ^
£
*->
2
o>
Q.
500
0 100
Note: VIN 1115. Packet #10
200 300 400
Time (seconds)
Figure 7
500
600
1984 LeBaron 2.2L Turbo, Original Catalyst
Bag #1, 75 deg. F FTP Cycle
Temperature and CO Efficiency
0.8
>.0.6
c
g>
"o
0.2
1,500
Catalyst
Temperature
CO Efficiency
1.000
500
o>
2
o)
CD
TJ.
CD
I
CD
D.
E
0 100
Note: VIN 0196. Packet #8
200 300 400
Time (seconds)
-11-
500
600
-------
Figure 8
Vehicle 0156, Original Catalyst
Bag#1,75deg. F FTP Cycle
Engine Out Emissions Cold/Hot Ratio
.O
OJ
DC
•*-•
O
X
TJ
O
0
IU
9
8
7
6
5
4
3
2
1
0
-
-
-
_
_
•
-
— ',
'''.*
,» /.« 1
1;. s\ / yx
• • • '". • / ::
'» N *l\ 1 / 1
- / *\-' '»*-"*- •'
1 »' " *.
1
1
1
1 A
** »*•• * /\
1 •
» '*
'•i / /": ;\
.-• • ; •/-. /•
; ; •-"-' »* ••• '•"-.
1, »
I 1 %1 1 H
0156
0 100
Note: VIN 0156, Packet #13
200 300
Time (seconds)
400
500
Figure 9
Vehicle 0192, Original Catalyst
Bag#1,75deg.F FTP Cycle
Engine Out Emissions Cold/Hot Ratio
CO
cc
i
O
O
IU
9
8
7
6
5
4
3
2
1
n
1
- >'
"
" 1
II
. 1 1
— 1 ! ) 1 1
1 1 '' to ' n
- 1 1 i, 1 II
.I.I11! 1 II
' V1! 1 1 I'
~ 1 ' 1 'i II '
- l-'Ht/ll ft
" ' 1 f\!w !'"^
5 \\ jiji^^ 'j ivv
i i
i
i
ii
i>
ii i
n n
n n
n n
i ii n i
,. ii nn
\ " . "u
1 H/ ^ MU .
. ll /\ H/l 1 |l l 'I
.s \ Vi / \ /Hi MM ' \
y s - ,' * V i ~" ' " ' ' v~-
"' w V \ i i ' i / i > /
1 1 1
0192
100
Note: VIN 0192. Packet #17
200 300
Time (seconds)
-12-
400
500
-------
Figure 10
Vehicle 1115, Original Catalyst
Bag#1,75deg.F FTP Cycle
Engine Out Emissions Cold/Hot Ratio
10
9
8
7
5
4
1115
I
I
I
'y '•••• ''••/ '""v'"
j • i
0 100
Note: VIN 1115. Packet #10
200 300
Time (seconds)
Figure 11
400
500
Vehicle 0196, Original Catalyst
Bag #1, 75 deg. F FTP Cycle
Engine Out Emissions Cold/Hot Ratio
10
9
B
7
08 6
CC
TO
8 4
3
2
1
0
0196
0 100
Note: VIN 0196. Packet #8
200 300
Time (seconds)
-13-
400
500
-------
1.0
0.0
0.8
0.7
g.08
§
'0 0.5
0.4
0.3
0.2
0.1
0.0
Figure 12
Average Catalyst Efficiency and Engine Out
Cold/Hot Ratio for Four Car EPA Data Sample
Cold Catalyst Efficiency
Hot Catalyst Efficiency
Engine Out C/H Ratio
1 i i i i i i i 11 i i i i i i i i t I i i i i i i i i i I i 1
100
200
300
400
500
g Speed vs. Time Trace
fi
600
Time (sec)
VEHSIME Code Development
Based on the substantially different warm—up characteristics for engines
and catalysts that were observed in the modal emissions data, it is
apparent that separate cold start adjustment equations for catalyst
warm-up and engine warm up are required. Figure 13 shows the warm-up
characteristics that were initially selected for coding into the VEHSIME
model.
The boldface lines in Figure 13 labeled "Catalyst Curve" and "Engine
Curve" represent cold start adjustments due to the catalyst and engine
warm—up, respectively. The catalyst curve is a simplified, step
function representation of the mean cold start catalyst efficiency of
the sample vehicles over time. The engine curve is a linear
representation of the mean engine out C/H ratio over time.
Mathematically, the catalyst warm-up and engine warm-up equations as a
function of time are:
CAT(t) - 0.80
--0
= 2.0
for t > 120 seconds
for t < 120
for t < 245 seconds
ENG(t) = -0.0049t
= 1.0
for 245 < t < 450
for t > 450
-14-
-------
Figure 13
Average Catalyst Efficiency and Engine Out
Cold/Hot Ratio for Four Car EPA Data Sample
Cold Catalyst Efficiency
Engine Out C/H Ratio
Speed vs. Time Trace
100
200
3OO
too
500
eoo
Time (sec)
The VEHSIME model was modified to include a user-invocable cold start
subroutine which applies the cold catalyst and engine equations to the
engine out emission levels of a fully-warmed engine map as follows:
COLD EMIS(t,RPM,load) = MAP EMIS(RPM,load) x (l-CAT(t)) x ENG(t)
Following some experimentation with the above algorithm, the catalyst
efficiency element of the algorithm was refined by substituting the
actual catalyst efficiency for each speed-load point computed from the
difference between the engine-out and tailpipe emission maps for the
individual engines.
Testing the Algorithm
In order to determine whether the cold start algorithm produced
reasonable results, VEHSIME simulations were run using engine-out and
tailpipe emissions maps available for three oxidation catalyst-equipped
engines (a 231 CID Buick, a 140 CID Ford and a 350 CID Chevrolet). A
4,000 pound chassis equipped with an automatic transmission was
simulated for the Buick engine. A 3,000 pound chassis with an automatic
-15-
-------
transmission was simulated for the Ford engine. A 5,000 pound chassis
with an automatic transmission was simulated for the Chevrolet engine.
Table 2 shows how the individual bag and composite FTP emissions for the
four vehicles tested by EPA compared to the emissions predicted by
VEHSIME for the three hypothetical vehicles.
As indicated by the data in Table 2, the predicted performance of the
hypothetical vehicle equipped with the 140 CID Ford engine indicates
that it is not capable of achieving warmed—up emission levels in the
range of those demonstrated by the four vehicles tested by EPA. If that
particular engine is excluded from consideration, the average emissions
of hypothetical vehicles using the other two engines are quite similar
to the average emissions of the vehicles tested by EPA. This similarity
was shown graphically in Figure 1 of the Summary section.
Table 2
Four-Car Sample FTP Emissions
Pollutant Vehicle
HC
CO
NOx
0192
1115
0156
0196
0192
1115
0156
0196
0192
1115
0156
0196
Bag 1 Bag 2
(grams) (grams)
4.32 1.24
3.15 1.21
3.33 0.24
2.02 0.35
31.96
37.07
33.86
22.53
3.00
3.81
4.47
4.48
10.21
22.56
3.46
4.89
0.72
2.64
2.97
2.93
Bag 3
(grams)
1.48
1.29
0.99
0.68
8.26
21.53
5.71
13.86
1.69
2.96
2.87
3.19
Bag1
(g/mi)
1.210
0.882
0.924
0.567
8.953
10.384
9.407
6.312
0.839
1.068
1.243
1.255
Bag 2
(q/mi)
0.323
0.314
0.061
0.090
2.659
5.847
0.892
1.255
0.187
0.685
0.764
0.753
Cold-72 Hot-72
Bag3 CompFTP (Bag 1+2) (Bag 3+2)
(g/mi) (g/mi) (g/mi) (g/mi)
0.417
0.365
0.274
0.187
2.325
6.068
1.587
3.847
0.476
0.834
0.797
0.886
0.525
0.440
0.298
0.214
3.821
6.770
2.837
2.997
0.396
0.795
0.870
0.890
0.741 0.363
0.581 0.333
0.476 0.164
0.316 0.137
5.623
7.951
4.976
3.656
0.496
0.860
0.992
0.988
2.463
5.879
1.223
2.500
0.321
0.747
0.779
0.816
VEHSIM Test Vehicle FTP Emissions
Bag 1 Bag 2 Bag 3
Pollutant Vehicle (grams) (grams) (grams)
HC
CO
Bu!ck231
Chevy 350
Ford 140
Buick231
Chevy 350
Ford 140
NOx Buick 231
Chevy 350
Ford 140
5.04 0.80 0.52
4.38 1.30 0.59
4.78 3.72 2.05
10.97
28.07
109.09
8.09
7.45
6.51
0.23
25.51
100.23
5.88
4.59
3.20
3.43
10.47
63.30
8.12
7.55
6.31
Bagl
(g/mi)
1.404
1.220
1.331
3.056
7.819
30.387
2.253
2.075
1.813
Bag 2
(g/mi)
0.205
0.332
0.951
0.059
6.524
25.634
1.504
1.174
0.818
Cold-72 Hot-72
Bag3 CompFTP (Bag 1+2) (Bag3+2)
(g/mi) (g/mi) (g/mi) (g/mi)
0.145
0.164
0.571
0.955
2.916
17.632
2.262
2.103
1.758
0.435
0.469
0.926
0.920
5.806
24.429
1.865
1.613
1.279
0.783
0.761
1.143
1.502
7.177
28.125
1.874
1.612
1.305
0.177
0.253
0.776
0.490
4.819
21.970
1.877
1.626
1.278
-16-
-------
Conclusions/Recommendations
The data sample on which this algorithm was based is very limited. With
only four vehicles, all equipped with fuel-injected, three-way catalyst
systems, the development of technology-specific relationships was not
possible. Another limitation of the cold start algorithm is that it is
being applied to engine maps for carbureted, oxidation catalyst equipped
vehicles.
Because of these limitations, and because of the extreme variation in
emissions from one vehicle to another, the most appropriate use of the
cold start algorithm at the present time may be as a means of developing
trip length "correction factors" to be applied to FTP-based emission
factors representative of the fleet of vehicles of interest. For
example, if the model predicts that CO emissions would be increased by
200% if the trip length is shortened from the length of the FTP to some
particular shorter distance, then that change in emission rate should be
applied to the FTP emissions rate using the best available emission
factors for the fleet in question, rather than the gram per mile values
predicted by VEHSIME.
In the future, the algorithm could be refined with the FTP Revision
Engine Mapping test data currently being collected by EPA. Under this
program, 30 low—mileage vehicles are being steady—state warm and cold
engine mapped. With a more robust sample, technology-specific effects
may be investigated. A refined algorithm could then be coded into
VEHSIME and compared against a representative sample of low—mileage FTP
data from surveillance testing or other testing programs.
-17-
-------
Appendix A
A Description of Sierra's
Vehicle Emissions Simulation Model
-------
A Description of Sierra's
Vehicle Emissions Simulation Model
VEHSIM, a Vehicle Simulation model, was originally developed by General
Motors in the early 1970's to produce a dynamic vehicle simulation
incorporating computations of instantaneous and cumulative fuel
consumption, engine speed, and engine load over any specified speed-time
profile. The model was expanded by the Department of Transportation in
the mid-1970's and the Environmental Protection Agency in the late
1970's to evaluate the effects of driving cycle changes on automobile
fuel economy and emission levels.
The VEHSIM program was modified by an EPA contractor to perform
simultaneous computations of emissions for HC, CO and NOx. This was
accomplished by writing a new program, VSIME, which utilizes the VEHSIM
program output for engine speed and torque time histories and engine
emission maps as inputs to calculate instantaneous and cumulative
emission rates over the driving cycle. The program output includes
emission quantities computed by VSIME and fuel consumption quantities
computed by VEHSIM.
Inputs to VEHSIM are organized into three categories:
- engine map for fuel consumption;
— driving cycle data; and
- vehicle configuration data.
The engine map is a matrix of fuel consumption rates and manifold vacuum
levels across a range of possible engine speed and load points. For
each discrete combination of speed (rpm) and load (in units of Ib-ft)
contained in the map, a fuel consumption rate (in pounds per hour) and
manifold vacuum (in inches of mercury below atmospheric) is provided.
Intermediate values are determined by an interpolation routine built
into the model.
The driving cycle file specifies the vehicle speed for each "segment" of
the cycle. Segments are nominally defined to be one second in length.
Cycle specifications are available for a variety of driving cycles,
including the LA4 driving cycle.
Vehicle configuration data characterize vehicle weight, frontal area,
aerodynamic drag coefficient, rolling resistance, drivetrain efficiency,
shift logic, fan losses, power steering losses, and air conditioning
losses. The shift logic is expressed for gear changes based on vehicle
speed and manifold vacuum.
A-l
-------
For a selected vehicle configuration, VEHSIM computes the engine speed,
load required to maintain the acceleration, and speed requirements set
for each segment of the specified driving cycle. The instantaneous fuel
consumption rate is determined by a double interpolation with respect to
speed and load within the engine map. The first two interpolations are
with respect to load within each of the relevant rpm settings. The
second interpolation is between the load values for each of the rpm
settings. The program employs a series of tests to determine whether a
vehicle has achieved the velocity required by a particular segment. The
length of the segment is extended as necessary in cases where the
selected combination of vehicle configuration and engine lacks the power
necessary to maintain the specified speed-time profile.
The outputs of VEHSIM include the following:
- cumulative distance (miles) and time (seconds);
- second-by-second and cumulative fuel consumption (pounds);
- second-by-second engine horsepower (hp) and torque (Ib-ft);
- second-by-second engine speed (rpm); and
— second—by—second manifold vacuum (inches of mercury).
Inputs to VSIME consist of the above outputs from VEHSIM plus engine
emission maps for HC, CO and NOx. Each engine emission map gives the
emission rate as a function of engine rotational speed and engine torque
(Ib-ft). The HC and NOx emission rates are input in units of grams per
hour. The CO emission rate is entered in units of 10 grams per hour.
For each segment of the driving cycle, the time duration is defined to
be one second. VSIME reads the load and the rpm values computed by
VEHSIM and uses that information to compute the instantaneous emission
rate. As with the fuel consumption calculation within VEHSIM, emission
rates are determined by double interpolation with respect to load and
rpm within each engine map. The VSIME program predicts HC, CO, and NOx
emissions for each second of the driving cycle. There are currently
seven different 1970s vintage engine maps in the format required by the
model; they span a size range from 91 to 350 cubic inches. For several
of the engines, the emissions map is available for both "engine-out" and
"tailpipe" emissions, after the exhaust passes through an oxidation
catalyst. Maps for late-model engines equipped with 3-way catalysts are
not yet incorporated in the model.
Currently, the model allows the user to select from seven different
chassis ranging from 2500 to 5000 pounds; however, it is possible to add
weight to these chassis through the user-selectible inputs. There are
also several different transmissions from which to choose.
During 1987, the emissions-prediction version of the model was recreated
by Sierra from a written description of the earlier EPA effort. (The
original version of the model which predicts emissions has been lost by
EPA.) Sierra's version of the model currently has over 4,000 lines of
A-2
-------
source code and comments. To distinguish Sierra's version of the model
from the original, we call the emissions model "VEHSIME".
One limitation of the original version of VEHSIME is that it only uses
data from warmed-up engines. As described in this report, a cold start
and warm-up routine has now been added. The cold start routine was
developed from second-by-second, pre-catalyst and after-catalyst modal
emissions data on vehicles tested by EPA. With the cold start routine,
the engine out emissions map is corrected based on the time since cold
start. HC and CO emissions are increased by a factor of almost 2:1 at
time zero. Stabilized engine out emissions are achieved at 450 seconds
after cold start. The cold start routine also determines when to begin
reducing the engine-out emissions by a catalyst efficiency factor
(computed from the difference between the engine out and tailpipe
emissions maps). Catalyst light-off is simulated at the point where the
cumulative fuel consumption reaches the level of fuel that would be
consumed during the first 2 minutes of operation on the LA4 cycle.
###
A-3
------- |