&EPA
United States
Environmental Protection
Agency
EPA-600/R-05/090a
August 2005
Heavy-Duty Diesel Vehicle
Modal Emission Model
(HDDV-MEM) Volume I:
Modal Emission Modeling
Framework
-------
-------
EPA-600/R-05/090a
August 2005
Heavy-Duty Diesel Vehicle Modal
Emission Model (HDDV-MEM)
Volume I: Modal Emission Modeling
Framework
by
Randall Guensler
Seungju Yoon
Chunxia Feng
Hainan Li
Jungwook Jun
School of Civil and Environmental Engineering
Georgia Institute of Technology
Atlanta, GA
Contract No: 4C-R022-NAEX
EPA Project Officer: Sue Kimbrough
U.S. Environmental Protection Agency
National Risk Management Research Laboratory
Air Pollution Prevention and Control Division
Research Triangle Park, NC 27711
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
Abstract
The research reported in this document outlines a proposed heavy-duty Diesel vehicle modal
emission modeling framework (HDDV-MEMF) for heavy-duty diesel-powered trucks and
buses. The heavy-duty vehicle modal modules being developed under this research effort,
although different from the structure within the motor vehicle emissions simulator (MOVES)
model, should be compatible with it. In the proposed HDDV-MEMF, emissions from
heavy-duty vehicles are predicted as a function of hours of on-road operation at specific engine
horsepower loads. Hence, the basic algorithms and matrix calculations in the new heavy-duty
diesel vehicle modeling framework should be transferable to MOVES. The specific
implementation approach employed by the research team to test the model in Atlanta is
somewhat different from other approaches in that an existing geographic information system
(GIS) based modeling tool is being adapted to the task. The new model implementation is
similar in general structure to the previous modal emission rate model known as the Mobile
Assessment System for Urban and Regional Evaluation (MEASURE) model.
Sponsored by the U.S. Environmental Protection Agency, this exploratory framework is
designed to be applied to a variety of policy assessments. The model can be used to evaluate
policies aimed at reducing the emission rates from heavy-duty vehicles as well as policies
designed to change the on-road operating characteristics to reduce emissions.
-------
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting
the Nation's land, air, and water resources. Under a mandate of national environmental laws,
the Agency strives to formulate and implement actions leading to a compatible balance
between human activities and the ability of natural systems to support and nurture life. To meet
this mandate, EPA's research program is providing data and technical support for solving
environmental problems today and building a science knowledge base necessary to manage
our ecological resources wisely, understand how pollutants affect our health, and prevent or
reduce environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency's center for
investigation of technological and management approaches for preventing and reducing risks
from pollution that threaten human health and the environment. The focus of the Laboratory's
research program is on methods and their cost-effectiveness for prevention and control of
pollution to air, land, water, and subsurface resources; protection of water quality in public
water systems; remediation of contaminated sites, sediments and ground water; prevention
and control of indoor air pollution; and restoration of ecosystems. NRMRL collaborates with
both public and private sector partners to foster technologies that reduce the cost of
compliance and to anticipate emerging problems. NRMRL's research provides solutions to
environmental problems by: developing and promoting technologies that protect and improve
the environment; advancing scientific and engineering information to support regulatory and
policy decisions; and providing the technical support and information transfer to ensure
implementation of environmental regulations and strategies at the national, state, and
community levels.
This publication has been produced as part of the Laboratory's strategic long-term research
plan. It is published and made available by EPA's Office of Research and Development to
assist the user community and to link researchers with their clients.
Sally Gutierrez, Director
National Risk Management Research Laboratory
in
-------
EPA Review Notice
This report has been peer and administratively reviewed by the U. S. Environmental Protection
Agency and approved for publication. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Information Service,
Springfield, Virginia 22161.
IV
-------
Table of Contents
Section Page
Abstract ii
Foreword iii
List of Figures viii
List of Tables ix
Acronyms x
Acknowledgments xiii
Executive Summary 1
Introduction 5
Modeling Parameters 7
Goals and Objectives 7
Pollutants Modeled 7
Approach 8
The Geographic Information System 13
GIS Spatial Analysis Framework 13
Infrastructure 14
Operating Environment 14
Subfleet Characterization and Traffic Volumes 17
Vehicle Technology Groups 17
Emitter Classifications 19
On-Road Traffic Volumes 19
Emissions Rate Differences 20
Transit Buses 21
Off-road Activity 21
Freight and Passenger Loads 23
Truck Loads 23
Transit Passenger Loads 25
On-Road Operating Characteristics 27
Engine Power Functions 29
Accessory Power Loss 31
Drive Train Power Loss 32
Mechanical Friction 32
Torque Converter Losses 32
Axle Bearing Friction 33
Inertial Losses 33
Driver Behavior 34
v
-------
Table of Contents (continued)
Section Page
Drive Train Power Loss Modeling Approach 35
Drive Train Efficiency 35
Rotational Moment of Inertia Losses 35
Axle Horsepower Relationships 36
System Monitors 36
Road Load Power Functions 37
Rolling Resistance Force (FR) 37
Gravitational Weight Force (Fw) 38
Aerodynamic Drag Force (FD) 39
Aerodynamic Lift Adjustment 40
Curve Resistance Force (Fc~) 40
Payload Inertial Resistance (Fp) 41
Available Tractive Force (FT) 41
Emission Rate Functions 43
Emitter Category 44
Commanded Fuel-Lean Operations 44
Correction Factors and Environmental Factors 45
Deterioration Rates 45
Temperature 46
Humidity 46
Altitude 46
Inventory Assembly and Model Output 49
Emission Matrix Calculations 49
Integration of Link-Based Emissions 50
Case Study of Two MARTA Transit Bus Routes 53
References 63
Bibliography 65
Appendix A Model Equation Parameters 67
Appendix B Potential Data Sources for the Heavy-Duty Vehicle Modal Emission
Modeling Framework 69
Appendix C Estimating Heavy-Duty Vehicle Miles Traveled 77
Heavy-Duty Vehicle Activity Data Sources 77
Highway Statistics Series 77
Vehicle Inventory and Use Survey 77
Highway Performance Monitoring System 78
Georgia Tech HDV/BUS Database 78
Heavy-Duty Vehicle VMT Estimation 79
Application in Road Load-Based Emissions Modeling 80
Appendix D Emission Testing Contacts 83
Appendix E Heavy-Duty Vehicle Emission Rate Data Sources and Applications 85
Emission Rate Data Available for Analysis 85
NVFEL (National Vehicle and Fuel Emissions Laboratory, EPA) 85
vi
-------
Table of Contents (concluded)
Section Page
NREL (National Renewable Energy Laboratory, DOE) 87
EERC (Engine and Emissions Research Center, WVU) 87
CIFER (Colorado Institute for Fuels and High Altitude Engine Research) 88
CAEC (Cleaire Advanced Emission Controls, LLC) 88
EMFAC2000 88
North Carolina State University 88
Application in the Load-Based Heavy-Duty Diesel Vehicle Modal Emission
Modeling Framework 89
Appendix F Modeling Approaches On-road Heavy-Duty Diesel Vehicle Oxides of
Nitrogen and Particulate Matter Emissions: PARTS, MOBILE6, EMFAC7G, and
EMFAC2000 91
PARTS 92
EMFAC7G 94
MOBILE6 94
EMFAC2000 94
vn
-------
List of Figures
Figure Page
1. Overview Schematic of the Proposed Model 10
2. Overview Schematic of the Proposed Model—Phase I Model 11
3. Overview Schematic of the Proposed Model—Phase II Model 12
4. Digitized Roadway Network—Example 15
5. X-Classification Scheme 17
6. VMT by EPA Classification—EPA Guidance vs Yoon Method 19
7. Heavy-Duty Vehicle Class Mapping Results 20
8. VMT Estimation Procedures 20
9. Horsepower Distributions by Class for Preliminary Surveys 24
10. Mean Horsepower by FHWA Truck Class 24
11. Engine Horsepower Distributions by Model Year Grouping 24
12. Class 9-13 Weight Distribution, Monroe County Weigh Station 25
13. Midday Time Period Weight Distributions, Class 9-13 Trucks 25
14. Speed/Acceleration Profile, Interstate Ramp, LOS D 27
15. Primary Elements in the Drive Train 29
16. Typical Aerodynamic Drag Coefficients 40
17. Example Emissions vs Axle-Horsepower 43
18. Stylized Plot of Axle-Horsepower vs. NOX Emissions (grams/second) 44
19. Graphic User Interface Developed for MEASURE 50
20. Example of Grid 8-9 AM Cell Emissions from MEASURE 51
21. Example MART A Transit Bus Routes on which Speed and Location
Data Collected 54
22. Speed-Acceleration Scatter Plots for Arterial Roads by Time of Day 55
23. Speed-Acceleration Scatter Plots for Local Roads by Time of Day 56
24. Speed-Acceleration Profiles for Arterial Roads by Time of Day 57
25. Speed-Acceleration Profiles for Local Roads by Time of Day 58
26. Example of a Speed-Acceleration Matrix (Arterial, Morning Peak Period) 59
27. Load Calculations for two Bus Routes with Different Grades and
Operating Profiles 61
B-l Typical Aerodynamic Drag Coefficients 75
B-2 Tire detail information 76
C-l Typical X-Scheme HDV Class Examples 78
C-2 Overall Process of HDV VMT Estimation 79
Vlll
-------
List of Tables
Table Page
1. Mapping of Functional Road Classifications for Modeling 16
2. Emissions Rates Differences between the X-Scheme and EPA Guidance 20
3. MARTA Bus Age Distribution 21
4. X-Scheme—HDV Reclassification Map among HDV and Truck Classification
Schemes 24
5. Typical Percent Increase in Effective Vehicle Weight (Excluding Trailer) 34
6. The Coefficient of Rolling Resistance Tire Parameters 38
7. Road Surface Coefficient 38
A-l. Model Equation Parameters 69
C-l Georgia statewide other 2-axle, 4-tire VMT Percentages 77
C-2 Georgia Statewide HDV VMT Fractions for the Class XI Conversion 78
C-3. EPA HDV VMT Fractions for Road Types 80
C-4. HDV and Bus VMT Fractions within the 20-County Atlanta Region 80
E-l. Heavy-Duty Vehicle HC Emission Rates for Use in MOBILE6 85
E-2. Heavy-Duty Vehicle NOX Emission Rates for Use in MOBILE6 86
E-3. Heavy-Duty Vehicle CO Emission Rates for Use in MOBILE6 86
E-4. Heavy Heavy-Duty Diesel Vehicle Emission Rates from NREL 87
E-5. Emissions Rates from CRC Project No. E-55/E-59 87
E-6. NYSDE Test Data used in EMFAC2000 88
E-7. CIFER Test Data Found in EMFAC2000 Document 88
E-8. Diesel HHDV Emissions Rates (g/mi) 88
F-l PM Emissions Estimation—Primary Model Components 91
F-2 NOX Emissions Estimation—Primary Model Components 92
IX
-------
Acronyms
AADT
ahp
ARB-HDETL
bhp
bhp-hr
CAEC
CARS
CBD
CE-CERT
CFR
CIFER
CMDL
CO
CPC
DEM
DOT
EDM
EERC
EMFAC2000
EMFAC2002
EMFAC7G
ESRI
FHWA
ft/s
ft/s2
FTP
g/mi
g/s
GDOT
GIS
GPS
GRTA
GVW
GVWR
HC
HDDV-MEMF
annual average daily traffic
axel horsepower
Air Resources Board Heavy Duty Emissions Testing Laboratory
brake-horsepower
brake-horsepower hour
Cleaire Advanced Emission Controls
California Air Resources Board
central business district
College of Engineering-Center for Environmental Research and
Technology (University of California at Riverside)
Code of Federal Regulations
Colorado Institute for Fuels and High Altitude Engine Research
Climate Monitoring Diagnostics Laboratory
carbon monoxide
Climate Prediction Center
digital elevation model
Department of Transportation
electronic distance measurer
Engine and Emissions Research Center
California Air Resource Board's mobile source emissions model
California's mobile source emissions model
California Air Resource Board's mobile source emissions model
Environmental Systems Research Institute, Inc.
Federal Highway Administration
feet per second
feet per second squared
federal test procedure
grams per mile
grams per second
Georgia Department of Transportation
geographic information system
Global Positioning System
Georgia Regional Transportation Authority
gross vehicle weight
gross vehicle weight rating
hydrocarbons
Heavy-Duty Diesel Vehicle Modal Emission Modeling Framework
x
-------
Acronyms (continued)
HDDV-MEM
HDDVs
HDGVs
HDV
HHDDVs
HPMS
HSS
km
Ibs
LOT
LDV
m/s
m/s2
MARTA
MEASURE
MHDDV
MOBILE6
MOVES
mph
mph/s
MTBE
MTPT
NCDC
NEPA
NOAA
NOX
NREL
NVFEL
NYSDEC
OBD
OTAQ
PARTS
PM25
ReFUEL
rpm
SAE
SAFD
SIP
TIUS
U.S. EPA
UDDS
VIUS
Heavy-Duty Diesel Vehicle Modal Emission Model
heavy-duty diesel vehicles
heavy-duty gasoline vehicles
heavy-duty vehicle
heavy heavy-duty diesel vehicles
Highway Performance Monitoring System
Highway Statistics Series
kilometer
pounds
light-duty truck
light-duty vehicle
meters per second
meters per second squared
Metropolitan Atlanta Rapid Transit Authority
Mobile Emissions Assessment System for Urban and Regional
Evaluation
medium heavy-duty diesel vehicle
EPA's mobile source emission factor model
motor vehicle emission simulator
miles per hour
miles per hour per second
methyl tertiary butyl ether
Multimodal Transportation Planning Tool
National Climatic Data Center
National Environmental Protection Act
National Oceanic and Atmospheric Administration
oxides of nitrogen
National Renewable Energy Laboratory
National Vehicle and Fuel Emissions Laboratory
New York State Department of Environmental Conservation and Energy
on-board diagnostic
EPA's Office of Transportation and Air Quality
EPA's Highway Vehicle Particulate Emission Factor Model
particulate matter less than 2.5 |im in aerodynamic diameter
Renewable Fuels and Lubricants Research Laboratory
revolutions per minute
Society of Automotive Engineers
speed/acceleration frequency distributions
State Implementation Plan
truck inventory and use survey
U.S. Environmental Protection Agency
urban dynamometer driving schedule
vehicle inventory and use survey
XI
-------
Acronyms (concluded)
VMT vehicle miles traveled
VOC volatile organic compounds
WVU West Virginia University
WVU-EERC West Virginia University - Engine and Emissions Research Center
ZML zero-mile level
xn
-------
ACKNOWLEDGMENTS
The authors wish to acknowledge the EPA Region 4 management and staff who have
contributed to the success of this project, specifically Beverly Bannister, Director, Air,
Pesticides and Toxics Management Division, Kay Prince, Chief, Air Planning Branch,
Thomas L. Baugh, and Dale Aspy.
Xlll
-------
Heavy-Duty Diesel Vehicle Model
xiv
-------
Emissions Modeling Framework
Executive Summary
The research reported in this document outlines a
proposed Heavy-Duty Diesel Vehicle Modal Emis-
sion Modeling Framework (HDDV-MEMF) for
heavy-duty diesel-powered trucks and buses. The
heavy-duty vehicle modal modules being developed
under this research, although different from the
structure within the Motor Vehicle Emissions Simu-
lator (MOVES) model, should be compatible with it.
MOVES is the next generation mobile source emis-
sions model being developed by EPA's Office of
Transportation and Air Quality (OTAQ) and will
replace the current MOBILE vehicle emission factor
model. MOBILE is used to calculate current and
future emission inventories of hydrocarbons (HC),
oxides of nitrogen (NOX), and carbon monoxide (CO)
from passenger cars, motorcycles, light- and heavy-
duty trucks at the national and local level. These
inventories are used to make decisions about air
pollution policy at the local, state, and national level.
Inventories based on MOBILE are also used to meet
the federal Clean Air Act's State Implementation
Plan (SIP) and transportation conformity require-
ments and are sometimes used to meet requirements
of the National Environmental Protection Act
(NEPA).
In the proposed HDDV-MEMF, emissions from
heavy-duty vehicles are predicted as a function of
hours of on-road operation at specific engine horse-
power loads. Hence, the basic algorithms and matrix
calculations in the new heavy-duty diesel vehicle
modeling framework should be transferable to
MOVES. The specific implementation approach
employed by the research team to test the model in
Atlanta is somewhat different in that an existing
modeling tool based on a geographic information
system (GIS) is being adapted to the task. The new
model implementation is similar in general structure
to previous modal emission model known as Mobile
Assessment System for Urban and Regional Evalua-
tion (MEASURE) model.
Historically EPA's mobile source emission rate
model (i.e., MOBILE) has produced emission rate
estimates based on average operating characteristics
and conditions (i.e., average "trip-based" modeling).
Average trip-based modeling refers to the use of
average in-use fleet emission factors. These emission
factors are developed on the basis of laboratory
dynamometer testing that simulates an average
vehicle trip. Although different driving cycles have
been developed over the years, conceptually dyna-
mometer testing is designed to obtain a "representa-
tive sample" of vehicle operations. Appropriate
emissions rates are developed and applied to an
activity rate such as vehicle-miles traveled to obtain
a set of average trip-based emission rates. The first
MOBILE model was developed in 1978. The current
generation of the model, MOBILE6, was released in
2002 after a lengthy update process. The chart below
shows that, even though vehicle speed varies
throughout a trip, the emissions are treated as a
Trip Cycle
(speed, operating conditions)
Average Trip-Based Emission Rate
(on-road portion)
Emission
Rate
1
-------
Heavy-Duty Diesel Vehicle Model
constant value with the average trip-based approach.
An additional advantage that the heavy-duty diesel
vehicle modal emission model (HDDV-MEM) has
over traditional modeling approaches is the imple-
mentation of modal modeling. "Modal" refers to the
type, or mode, of engine operation that a vehicle may
be in at any given point in time. Important vehicle
operating modes include engine start, engine idle, hot
stabilized operation (on-road operation), enrichment
conditions (influenced by high acceleration and
power demand), and hot soak evaporation. Appropri-
ate emission rates are developed and applied to the
various vehicle modes of operation to obtain a set of
modal emissions. The chart below shows a stylistic
representation of how emissions could vary by mode
for a typical vehicle trip.
Engine Start
Emission
Rate
Hot Stabilized
and Running Losses
Time (seconds)
Historically, emission rate or emission models have
not been implemented in a GIS. This has been due in
part to the magnitude of the data required by a GIS,
perception and reality of operating GIS software, and
the level-of-knowledge required by end-users to
manipulate data in a GIS. In the last 10 years, GIS
data have become more available in both quantity and
quality. Availability of a GIS in a Windows operating
system environment, availability of low-cost com-
puter hardware, and the development and implemen-
tation of easy to use GIS tools (in a Windows envi-
ronment) has made implementation of GIS-based
emissions models more practical.
Models, such as MOBILE6, do not calculate emis-
sions due to road grade effects—increased emissions
due to accelerations going uphill. In addition, spatial
resolution of emissions is misrepresented. For exam-
ple, engine start emissions using earlier modeling
regimes did not allocate emissions to the appropriate
residential or commercial/industrial locations. Typi-
cally, engine starts would be included in the grand
total of highway vehicle emissions and would be
allocated a given county based on population or some
other surrogate. With a GIS, engine start emissions
may be calculated and assigned on a sub-county basis
using a combination of factors including census block
group population, property (parcel) data, trip produc-
tion/attraction (work, schools, etc.) data.
The use of a GIS-based approach allows for the
visualization of such critical phenomena as real-
world locations and magnitude of emissions from
vehicles during peak commuting hours on major
roadways. One of the more difficult highway vehicle
emissions modeling questions to answer is how to
locate the emissions—where geographically are the
emissions coming from? In a large metropolitan area,
the highway vehicle emissions occur throughout the
city and throughout the day.
There are several key features that make the load-
based modeling approach more appropriate for
emissions prediction than current emission rate
modeling tools:
• Modal models take into account all of the factors
in the heavy-duty vehicle operation environment
that affect emissions, such as vehicle age, engine
type, transmission type, fuel type, on-road driv-
ing conditions, and roadway characteristics.
• The statistic methodology approach avoids
extrapolation with correction factors beyond
ranges under which test data were collected,
significantly improving prediction accuracy.
• Modal models are easily verified, calibrated, and
improved. Second-by-second vehicle operations
can be readily collected in field studies, and new
heavy-duty vehicle emission rate monitoring
equipment can be deployed to verify instanta-
neous emission rates. Field test results can be
used to calibrate the parameters of the load-based
model accordingly.
-------
Emissions Modeling Framework
Phase I of the project has been to develop the model-
ing concepts and construct a working model. This
development effort is described more fully in the
body of this report (Volume I and II). Phase II of the
project includes: (1) collection of available emission
testing data to support proposed load-based emission
modeling approaches, (2) development and applica-
tion of analytical and statistical methods to support
the proposed emission modeling approaches, (3)
evaluation and refinement of emissions modeling
approaches based upon available data, (4) the devel-
opment of effectively executable model source codes,
and (5) model sensitivity analysis through the itera-
tive evaluation of the impacts of structured external
data files (with the goal of improving model effi-
ciency without reducing model resolution and predic-
tive capabilities).
-------
Heavy-Duty Diesel Vehicle Model
-------
Emissions Modeling Framework
Introduction
Heavy-duty vehicles are a major source of pollutant
emissions in metropolitan areas. The relative impor-
tance of reducing emissions from heavy-duty diesel
vehicles (HDDVs) has become critical in solving, for
example, Atlanta's air pollution problems. HDDVs
constitute a small portion of the fleet but emit a
disproportionately large amount of oxides of nitrogen
and paniculate matter. As light-duty vehicle (LDV)
emission rates continue to decline, heavy-duty
vehicles (HDVs) are increasingly looked toward as a
source for additional emission reductions. Diesel
vehicles have also been classified as the largest
source of mobile source air toxics.
Emissions from light-duty cars and small trucks have
been stringently controlled for many years. Signifi-
cant research efforts has been conducted to develop
models capable of predicting travel demand, on-road
vehicle operating conditions, and the emissions
generated by these vehicles. Effective control strate-
gies for LDVs have been proposed and implemented
through analysis of such detailed research data.
However, emission standards for HDDVs have only
recently become more stringent, and very little re-
search into the driving activities and patterns, espe-
cially in a local area, has been conducted. Therefore,
it is difficult to assess the cost-effectiveness of HDV
emission reduction strategies.
As it stands today, the emission rate models for
HDVs still require significant improvement. Major
modeling defects for the heavy-duty components
within the U.S. Environmental Protection Agency's
(U.S. EPA's) MOBILE models have been widely
recognized for more than 10 years (Guensler et al.,
1991). The current model used by 49 states and the
federal agencies is the MOBILE6 model, developed
by the EPA. The release of MOBILE6 contained only
marginal improvements to the correction factors
employed in the model and has not resolved the
fundamental flaws inherent in the modeling approach.
Under the current MOBILE6 modeling regime, HDV
emissions are essentially predicted as a function of
miles traveled and average speed. A major overhaul
of the basic modeling approach in the MOBILE6
model is warranted.
Since 2000, the EPA has been developing a new
generation mobile source emissions model (motor
vehicle emission simulator, MOVES) based on the
specific horsepower requirements (horsepower to
weight ratio) of vehicles. Specific horsepower is a
function of speed, acceleration, and road grade. In
general, a vehicle at higher speed, harder accelera-
tion, and steeper road grade requires more specific
horsepower for the vehicle to overcome drag forces
and to move.
The research reported in this document outlines
HDDV-MEMF, a proposed HDDV modal emission
modeling framework for heavy-duty diesel-powered
trucks and buses. The HDV modal modules being
developed under this research, although different
from the structure within MOVES, should be compat-
ible with it. In the proposed HDDV-MEMF, emis-
sions from heavy-duty vehicles are predicted as a
function of hours of on-road operation at specific
engine horsepower loads. Hence, the basic algorithms
and matrix calculations in the new heavy-duty diesel
vehicle modeling framework should be transferable
to MOVES. The specific implementation approach
employed by the research team to test the model in
Atlanta is somewhat different from other approaches
in that an existing GIS-based modeling tool is being
-------
Heavy-Duty Diesel Vehicle Model
adapted to the task. The new model implementation
is similar in general structure to the previous modal
emission model known as MEASURE (Bachman, et
al, 2000; Bachman, 1997).
Sponsored by the EPA, this exploratory framework is
designed to be applied to a variety of policy assess-
ments. The model can be used to evaluate policies
aimed at reducing emissions from HDVs as well as
policies designed to change the on-road operating
characteristics to reduce emissions. Once the model-
ing framework is complete and field tested in Atlanta,
the model framework will be applicable to other
major metropolitan areas, provided that the required
data inputs are assembled for these areas. However,
even in the absence of metro-specific data, the default
parameters associated with Atlanta can be employed.
-------
Emissions Modeling Framework
Modeling Parameters
Goals and Objectives
The goal of the proposed research is to develop a
load-based emission model for on-road heavy-duty
diesel trucks and transit vehicles within a GIS frame-
work. The HDDV emissions module will be devel-
oped in the same general framework used to develop
the MEASURE model (Bachman et al., 2000).
Refined HDDV activity estimates will be combined
with improved emission rates to produce an HDDV
emissions module that can be evaluated for inclusion
into the next generation EPA emission model known
as MOVES. Thus, both models are expected to be
able to more accurately estimate on-road HDDV
emissions than current modeling techniques. Once
integrated, the improved modeling tools will enable
states to develop and implement more effective
HDDV emission reduction strategies.
The GIS-based approach provides the ability to
spatially and temporally evaluate the criteria pollutant
emissions impacts for a variety of freight and transit-
oriented policies even when vehicle operating condi-
tions change over time. The model will:
• Operate within a GIS framework, allowing users
to evaluate changes in the spatial and temporal
distribution of public transit emissions;
• Support alternatives evaluation at the regional as
well as microscale levels;
• Allow assessment of the emissions impacts of
changes in truck and transit technology purchases
(such as vehicle size, engine classification,
dedicated fuel type, etc.) and fleet deployment;
• Support evaluation of emissions effects resulting
from changes in transit or delivery routes, pas-
senger and freight loading, and operating duty
cycles associated with congestion or high speed
operations;
• Employ a graphic user interface; and
• Provide transferability to any maj or urban area in
the country.
Pollutants Modeled
The model is designed to directly predict emissions
on a load basis for three criteria pollutants: CO, NOX,
and volatile organic compounds (VOCs). Although
the model emphasis is on HDDVs and their emis-
sions, heavy-duty gasoline vehicles (HDGVs) are
included in the model. For this reason, emissions of
benzene, methyl tertiary butyl ether (MTBE), 1,3
butadiene, formaldehyde, acetaldehyde, and acrolein
are included in the model. Speciation factors are
applied to the VOC emissions to predict emissions of
benzene, MTBE, 1,3 butadiene, formaldehyde, acet-
aldehyde, and acrolein. Although recent research
indicates that speciation factors vary as a function of
engine load, the data are still too sparse to implement
load-related speciation profiles at this time. However,
as information continues to improve, load-based
speciation factors can be handled by the model. It is
simply a matter of creating separate emission rate
matrices (gram per second) as a function of brake-
horsepower (bhp) load in the calculation methodol-
ogy rather than applying a uniform speciation faction
after VOC emissions are totaled.
Emissions of exhaust particulate matter less than 2.5
|im in aerodynamic diameter (PM25) are also pre-
dicted directly on a load basis, with simple speciation
factors applied to the total PM2 5 to predict emissions
of sulfate, organic carbon, elemental carbon, gaseous
PM, and lead. Fricti on-based PM25 emissions for tire
wear and brake wear are also integrated as a function
of miles traveled. As noted with air toxics, separate
PM emission rate functions can be implemented for
7
-------
Heavy-Duty Diesel Vehicle Model
each sub-species as data become available. Separate
emission rate matrices (gram per second) as a func-
tion of bhp load in the calculation methodology can
be developed as new laboratory testing data are made
available.
Carbon dioxide emissions are implemented through
the integration of brake-specific fuel consumption
estimates, with all carbon in the fuel assumed to
convert to carbon dioxide. As carbon dioxide emis-
sion rates in grams per second become available for
analysis, variable emission rates as a function of
instantaneous vehicle power demand can be imple-
mented in the Phase II model.
Sulfur dioxide and ammonia (applicable to HDGVs
only) emission rate testing results are sparse. As such,
implementation of modal algorithms will not likely
occur in Phase I modeling. Available gram per brake-
horsepower-hour emission rates will be collected
from the literature and applied.
Modeling Approach
The proposed model for predicting transit emission is
designed for transportation infrastructure implemen-
tation on a link-by-link basis. Although the modeling
routines are actually amenable to implementation on
a vehicle-by-vehicle basis, the large number of
vehicles operating on infrastructure links precludes
practical application of the model in this manner. As
such, the model framework capitalizes upon previous
experience gained in development of the MEASURE
modeling framework in which vehicle technology
groups were employed. However, whereas the MEA-
SURE model employed load surrogates for the
implementation of a light-duty modal modeling
regime, this new modeling framework directly
predicts heavy-duty vehicle operating loads and uses
these load predictions directly in the emission predic-
tion process.
Vehicle technology groups are employed more
extensively in the HDDV-MEMF than they were in
MEASURE. In MEASURE, technology groups were
relatively independent of vehicle configuration and
based solely upon baseline laboratory emission test
results. That is, groups of vehicles that behaved
similarly to each other on the baseline tests, and
responded similarly to alternative tests, were grouped
together into a vehicle technology. In the heavy-duty
vehicle world, drive train (engine, transmission,
differential, and tires) design, truck and trailer physi-
cal configuration, and the cargo loads that they carry
all affect on-road operating loads. Hence, in the
HDDV-MEMF, technology groups must relate to the
subfleet composition (measurable on-road vehicle
classifications), the drive train characteristics, as well
as the performance of the various engine classes in
laboratory testing. This type of model is known as a
modal model because it directly predicts second-by-
second emissions from any on-road driving mode.
For each technology group, the model predicts engine
power demand in response to inertial load, grade
load, road friction, accessory load (e.g., air condition-
ing usage), given the distributions of second-by-
second operating modes for the on-road vehicles.
Such on-road activity can be developed through
empirical observation, using laser guns or instru-
mented fleets, collecting such data on a second-by-
second basis.
Emissions rates for each engine and vehicle family
(engine manufacturer, displacement, certification
family, drivetrain, fuel delivery system, emission
control system) are first established in grams per
brake-horsepower hour (from standard engine dyna-
mometer certification data or from on-road emission
rate data when available). Basic engine power equa-
tions are employed to predict engine load (brake-
horsepower) for every second of operation (brake-
horsepower-sec) as a function of environmental and
operating conditions for the specific vehicle technol-
ogy. Emissions are then determined by integrating
predicted emissions over time (grams per brake-
horsepower-hour multiplied by brake-horsepower-
hour). This research will include initial model devel-
opment and supplemental data collection (Phase I)
and ongoing data collection, model refinement,
calibration, and testing (Phase II). The project will
-------
Emissions Modeling Framework
culminate in the application of the modeling system
to a case study in Atlanta, Georgia.
There are several key features that make the load
based modeling approach more appropriate for
emissions prediction than current emission rate
modeling tools:
• Modal models take into account all of the factors
in the heavy-duty vehicle operation environment
that affect emissions, such as vehicle age, engine
type, transmission type, fuel type, on-road driv-
ing conditions, and roadway characteristics;
• The statistic methodology approach avoids
extrapolation with correction factors beyond
ranges under which test data were collected,
significantly improving prediction accuracy;
• Modal models are easily verified, calibrated, and
improved. Second-by-second vehicle operations
can be readily collected in field studies, and new
heavy-duty vehicle emission rate monitoring
equipment can be deployed to verify instanta-
neous emission rates. Field test results can be
used to calibrate the parameters of the load-based
model accordingly.
As outlined above, model inputs include load related
parameters and grams per brake-horsepower-hour
emissions rates under different power demand situa-
tions. In the Phase I model, constant grams per brake-
horsepower-hour emissions rates are derived from the
EPA laboratory test data and will be corrected based
on altitude, temperature, humidity and whether the
vehicle is assigned to a normal-emitter or high-
emitter category (using a probability function). Load
parameters will be determined by: (1) vehicle tech-
nology, including vehicle type, make, model year,
engine type, transmission type, frontal area, drag
coefficient, rolling resistance, vehicle maintenance
history, and so forth; (2) loading factors, including
passenger load and freight load; (3) roadway charac-
teristics, including road grade and, possibly, pave-
ment surface roughness; (4) and inertial load parame-
ters based on the vehicle speed and acceleration
profile and the environmental conditions.
As discussed in the model overview section, once the
brake-horsepower demands are quantified, engine
dynamometer emission rate data (grams per brake-
horsepower-hour) can be used in the emissions
modeling regime. Note, however, as new emission
rate data are collected via chassis dynamometer
testing or through actual on-road testing, in which
axle horsepower loads are measured concurrently
with emission rates (grams per axle-horsepower-
hour), these emission rates can be employed directly,
without the need for estimating additional power
losses associated with drive train and accessory
losses.
The remainder of the report describes the details of
the modeling framework. The GIS modeling regime
is outlined first. Figures 1, 2, and 3 illustrate the
combined model regime, the Phase I implementation,
and the Phase II implementation. The implementation
figures provide blue boxes around each of the major
modeling elements. Each major component of the
model illustrated in these figures is described in a
separate section of this report. The roadway infra-
structure section outlines the detailed information
needed for the link-by-link model implementation.
Because temperature, humidity, altitude, and wind
speed affect vehicle loads and emission rates, the
tracking of these factors is outlined in one report
section. Vehicle activity must be estimated on each
link for a variety of heavy-duty vehicle subfleets.
The methods proposed for developing the vehicle
subsets and for estimating their activity are described
in the Subfleet Characterization and Traffic Volumes
section. This section also refers to a number of
technical appendices. Freight and passenger load
estimation are described in one section, and on-road
operating characteristics (speed and acceleration
profiles) are described in another. The most complex
section of the report is the description of the engine
power functions. Emission rates are significantly
different for Phase I and Phase II modeling
approaches, and the figures indicate that there are
separate modules associated with each. The emission
rate functions section describes the approach differ-
-------
Heavy-Duty Diesel Vehicle Model
ences and outlines the methods proposed for imple-
menting correction factors and more complex
time-series relationships. The matrix calculation
methods are outlined in the inventory and assembly
section. Finally, the report provides a case study
citation of the basic power approach to estimation of
emisisons from two monitored transit routes in
Atlanta.
Heavy-Duty Diesel Vehicle Modal Modeling Framework
Driver Behavior
(throttle and shift pattern
Knginc Parameters
(RPM fuel rate, clc.)
Hnviromttent:
•Altitude
•Temperature
•Wind fields
•fkmiiditv
Load Parameters
•V: \ehicle speed
•A: acceleration
•0: grade
•(.': drag coefficient
•M: \ chicle mass
•F; rolling resistance
•S; cross section area
•p: air density
•W: wind speed
Driving Trace (sec-by-see)
Speed-'Acceleration Profi le
Basic Emission Rate
l;unctions
(g.-'bhp-hr)
B!:R Correction 1'actors:
Altitude, 'I'eniperature,
Deterioration, Humidits
Ha/ard Models for
C'omponent ! ailure
Road%vav Infhtslructure
Kxternal Sensors
linginc Noise-1 Vibration,
Wheel Vibration,
Tire Pressure
Gucnsler, et al., 2004
Figure 1. Overview Schematic of the Proposed Model.
10
-------
CQ
C
10
O
0
o
IT
0
3
03
^-t-
o'
O
0
o
T3
O
C/>
0
Q.
O
Q.
0_
TJ
IT
03
C/>
0
O
Q.
0
HDDV Modal Modeling Framework - Phase I
Environment:
•Altitude
•Temperature
•Wind fields
•Humidity
Load Parameters
•V: vehicle speed
•A: acceleration
•0: grade
•C: drag coefficient
•M: vehicle mass
•E: rolling resistance
•S: cross section area
•p: air density
•W: wind speed
Accrued VMT
Accrued VMT
A.
B.
C.
D.
E.
f.
C.
L.
c,
Normal/High !• miller
I'raclions
Infrastructure
Operating Environment
Volumes and Subtteets
Freight/Passenger Loads
Onroad Operations
Engine Power Functions
Base Emission Rate Functions
Inventory Assembly
Environment:
•Altitude
•Temperature
•Mumidilv
Basic Emission Rate
functions
(g/bhp-hr)
G
BER Correction Eactors:
Altitude, Temperature.
Deterioration, Humidity
m
3
'
(A
o'
O
Q.
SL
5'
(Q
D)
3
(D
O
Guensler, ct al., 2004
-------
(Q
C
CO
O
U)
o
3
03
l—t-
a
o
o
T3
o
V)
0
Q.
o
Q.
0_
TJ
V)
0
o
Q.
0
HDDV Modal Modeling Framework - Phase II
Driver Behavior
(Ihrottlc and shirt patterns)
Kngine Parameters
(RPM, fuel rate, etc.)
H. Modal Emission Rates
I. Emitter Characterization
J. Driver Behavior
K. System Monitors
L. Inventory Assembly
Environment:
•Altitude
•Temperature
•Wind fields
•Humidity
Driving Trace (sec-by-sec)
Speed/Acceleration Profile
Accrued VMT
Accrued VI IT
Maintenance Histor\
Ha/ard Models lor
Component l-'ailurc
Modal
Emission Rate
(•"unctions
&'s=Hbhp)
Road Load Power
Hxlernal Sensors K
Bngine Noise/Vibration,
Wheel Vibration,
'('ire Pressure
Koadvvas' Inirastructtire
Load Parameters
•V: vehicle speed
•a: acceleration
•0: grade
•C: drag coefficient
•M: vehicle mass
•F: rolling resistance
•S: cross section area
•p: air density
•W: wind speed
Guensler, et al., 2004
(D
D)
6
P*
D
(D
O
Q.
(D
-------
Emissions Modeling Framework
The Geographic Information System
GIS Spatial Analysis Framework
Historically, emission rate or emission models have
not been implemented in a GIS. This has been due, in
part, to the magnitude of the data required by a GIS,
perception and reality of operating GIS software, and
the level-of-knowledge required by end-users to ma-
nipulate data in a GIS. In the last 10 years, GIS data
have become more available in both quantity and
quality, while the availability of a GIS in a Windows
operating system environment, availability of low-
cost computer hardware, and the development and
implementation of easy to use GIS tools (in a Win-
dows environment) has made implementation of
GIS-based emissions models more practical.
Models, such as MOBILE6, do not calculate emis-
sions due to road grade effects—increased emissions
due to accelerations going uphill. In addition, spatial
resolution of emissions is misrepresented. For exam-
ple, engine start emissions using earlier modeling
systems did not allocate emissions to the appropriate
residential or commercial/industrial locations. Typi-
cally, engine starts would be included in the grand
total of highway vehicle emissions and would be
allocated a given county based on population or some
other surrogate. With a GIS, engine start emissions
may be calculated and assigned on a sub-county basis
using a combination of factors including census block
group population, property (parcel) data, trip produc-
tion/attraction (work, schools, etc.) data.
The use of a GIS-based approach allows for the
visualization of such critical phenomena as real-
world locations and magnitude of emissions from
vehicles during peak commuting hours on major
roadways. One of the more difficult highway vehicle
emissions modeling questions to answer is how to
locate the emissions—that is to say, where (geo-
graphically) are the emissions coming from? In a
large metropolitan area, the highway vehicle emis-
sions occur throughout the city and throughout the
day.
The spatial analysis framework for the heavy-duty
diesel modal emission model is an implementation in
ArcGIS. The GIS system is essentially a spatial
database that tracks the physical location, spatial
boundaries (shapes), and associated attributes (physi-
cal or performance characteristics) of the modeled
elements. The GIS system contains the land use and
roadway infrastructure data that are used in a variety
of the emissions calculations. Variables included in
the proposed research model are those required for
use in the quantification of vehicle activity or the
calculation of vehicle emissions. Calculations are
performed for each roadway link, (i.e., on a link-by-
link basis). Hence, the spatial transportation infra-
structure must be described with the applicable
link-based variables. Modeling is also performed on
an hourly basis, so attributes are also maintained for
each hour. Given the functional form of the load-
based model, the elements listed below are tracked in
the GIS system.
Land Use:
• U.S. Census block boundaries,
• Parcel level land use boundaries (to identify truck
activity locations),
• Traffic analysis zone boundaries (from the re-
gional travel demand forecasting model), and
• Grid cell boundaries (defined by user for regional
air quality modeling).
Roadway Elements:
• Travel demand forecasting network link identi-
13
-------
Heavy-Duty Diesel Vehicle Model
fication,
• Link x, y coordinates,
• Roadway length,
• Roadway classification,
• Number of lanes,
• Speed limit,
• Road surface material (e.g., concrete vs asphalt),
and
• Grade distribution and average grade.
Temporal Variable:
• Hour of the day (0-24).
Traffic volumes:
• HD diesel vehicle traffic flow (vehicles by link
by hour).
On-road Fleet Characterization:
• HD vehicle classification;
- Model year,
- Engine size, and
- Vehicle weight,
• Vehicle configuration group;
• Drive train technology group, and
• Emission control system.
On-road Operating Conditions:
• Speed/acceleration profile.
Environmental Conditions:
• Temperature,
• Humidity,
• Altitude, and
• Wind speed and direction.
The spatial domain is currently the 13-county Atlanta
metropolitan area, but will be expanded to the new
20-county nonattainment area, which is designated as
an 8-hr ozone nonattainment area, over the next year.
The use of each element is described later in the
modeling framework documentation associated with
that element.
Infrastructure
The roadway elements included in the model are:
• Travel demand forecasting network link identifi-
cation,
• State DOT/FHWA roadway classification,
• Link x, y coordinates,
• Roadway length,
• Roadway classification,
• Number of lanes,
• Speed limit,
• Surface material (e.g., concrete vs asphalt), and
• Grade distribution and average grade
The travel demand forecasting link identification
provides a unique roadway ID for use in all calcula-
tions. Link coordinates provide the spatial location of
each link. Because the demand modeling framework
is a simplified version of the actual on-road system,
each link is represented by a straight line (Figure 4).
To ensure that the correct distance and travel time is
included in all calculations, the actual on-road length
of each roadway link is included as a link attribute.
Roadway classification, number of lanes, and speed
limits are also included for use in later development
of applicable on-road operating conditions
(speed/acceleration matrices) for each link. Surface
materials are tracked for use in roadway friction
calculations. Finally, grade is tracked for use in the
load-based gravitational resistance calculation.
The four roadway classification definitions that are
employed in the modeling framework are freeways,
arterial/collector, local roads, and ramps. Both the
Federal Highway Administration (FHWA) and the
Georgia State Department of Transportation's
(DOTs) Multimodal Transportation Planning Tool
(MTPT) employ more refined roadway classification
systems, but each one is significantly different from
the other. A mapping system was developed for the
FHWA and State DOT roadway classifications
(Table 1) to assign each road link to the four classes
(Guensler, et al., 2004).
Operating Environment
Each grid cell used for aggregating emissions (1 km
x 1 km, or higher level) is assigned operating envi-
ronment attributes for temperature, humidity, altitude,
and wind speed and direction. The environmental
attributes are used in load calculations (temperature
and wind speed) and in emission rate adjustments
(temperature, humidity, and altitude). The environ-
mental conditions will be established for any given
14
-------
Emissions Modeling Framework
Land Use Overlay
Infrastructure Links (Red and Blue)
\ .-
Traffic Volume Example (MEASURE LDVs)
Figure 4. Digitized Roadway Network—Example (Bachman, et al., 2000;
Bachman, 1997).
modeling scenario. For example, summertime tem-
perature fields would be employed in estimating the
emissions contribution of HDDVs for ozone model-
ing. Particulate matter impact simulation studies
might model hourly emissions using the previous
year's hourly temperature fields (24 hours times 365
days). The spatial and temporal distributions of
temperature and humidity can be derived from
ambient monitoring data. Altitude (low vs high) is a
single variable for the entire modeling run. Hourly
wind speed and direction from meteorology monitor-
ing networks will be employed in the Phase II model
and used for establishing effective velocity in on-road
wind drag force calculations, provided that the
emissions effect is significant. Variables included in
the model are outlined in Appendix A and data
sources are outlined in Appendix B.
15
-------
Heavy-Duty Diesel Vehicle Model
Table 1. Mapping of Functional Road Classifications for Modeling3.
MOBILE6.2
Categories
FHWA Corresponding Categories
GA MTPT Corresponding Categories
Freeway
Arterial/Collector
Local
Ramp
Interstate Rural, Urban (FHWA Class
1,11)
Principal Arterial Rural (FHWA Class
2)
Other Freeways & Exp. Urban (FHWA
Class 12)
Minor Arterial Rural, Urban (FHWA
Class 6, 16)
Major Collector Rural (FHWA Class
7)
Other Principal Arterial Urban (FHWA
Class 14)
Minor Collector Rural (FHWA Class
8)
Local Rural, Local Urban (FHWA
Class 9, 19)
Collector Urban (FHWA Class 17)
None
Interstate Rural, Urban (FHWA Class 1, 11)
Urban freeway and expressway (FHWA Class
12)
Principal Arterial Rural (FHWA Class 2)
Urban principal arterial (FHWA Class 14)
Minor Arterial Rural, Urban (FHWA Class 6,
16)
Major Collector Rural (FHWA Class 7)
NFA Minor Collector Street with speed limit >
40 mph (FHWA Class 8)
Collector Urban (FHWA Class 17)
Non-Ramp Local Rural, Local Urban with
speed limit > 40 mph (FHWA Class 9, 19)
NFA Minor Collector Street with speed limit =
40 mph) (FHWA Class 8)
Non-Ramp Local Rural, Local Urban with
speed limit = 40 mph (FHWA Class 9, 19)
Ramps designated at Local Rural, Local Urban
(FHWA Class 9, 19) but defined as an
RCLINK code 6 (Ramp/ Interchange)in the
Georgia database
1 Source: Guensler et al., 2004
16
-------
Emissions Modeling Framework
Subfleet Characterization and Traffic Volumes
HDV activity must be estimated for each link on the
roadway system. Rather than tracking individual
vehicles (given the tremendous number of link-based
calculations that would need to be performed), the
model employs the concept of technology groups,
which are sets of vehicles that behave similarly to
each other and are assumed to follow the same load
and emissions relationships. This allows the emis-
sions for hundreds to vehicles to be predicted in a
single series of calculations. Hence, for each roadway
link, the activity of each technology group must be
provided. The unit of vehicle activity for each road-
way link is vehicle flow (vehicles per link per hour).
On the vehicle activity side, the modeling framework
will be flexible. Users can apply a default truck
fraction for each link to the link-based traffic vol-
umes output from traditional 4-step travel demand
models. Many regions apply a simple 0%, 1%, 2%, or
5% value to the travel demand predictions to estimate
the total number of vehicles operating on a link. Such
estimations are traditionally based upon truck count
observations and do not provide any insight into trip
generation for truck activity. Unfortunately, there are
no metropolitan areas that are currently operating
models that estimate trip generation, distribution, and
route choice for HDV activity (this would be a very
expensive, data intensive, and resource intensive
proposition). Given that there is no current means to
integrate a predictive model into the current modeling
regimes, it becomes imperative to provide the flexi-
bility for regions to integrate the results of special
research studies into the model.
Vehicle Technology Groups
The concept of vehicle technology groups is to
identify and track subsets of vehicles that have
similar on-road load response and similar laboratory
emissions performance. Vehicles within each tech-
nology group should respond similarly to change in
operating conditions (with respect to increased or
decreased load) and perform similarly to each other
on baseline emission tests. The basic premise is that
vehicles in the same HDV class, employing a similar
drive train, and of the same size and shape will
respond similarly with respect to load estimation.
There is also an important practical consideration in
establishing vehicle technology groups; researchers
need to be able to identify these vehicles in the field
during traffic counting exercises. The starting point
for technology group criteria is a visual classification
scheme. Justification for the visual classification
scheme is provided in a separate publication (Yoon
et al., 2004a).
• The vehicle visual classification X-Scheme
(Yoon et al., 2004a), from a small single axle
truck to a large multiple trailer truck, is illus-
trated in Figure 5.
X Class EPA Class
Typical Vehicles
X1
X2
X3
HDV2b, HDV3,
HDV4, HDV5,
HDV6, HDV7
HDVSa
HDVSb
Figure 5. X-Classification Scheme (Yoon et al.,
2004a).
17
-------
Heavy-Duty Diesel Vehicle Model
Vehicle configuration (bobtail, tanker, single-
unit, trailer, double trailer, flat-top, etc.) has a
significant impact on engine load at high speeds
through the aerodynamic drag load. Therefore,
body configuration is also employed as a technol-
ogy group criterion.
Three engine size classifications are associated
with both engine design and certification, and
these will be used in technology group criteria: of
light heavy-duty, medium heavy-duty, and heavy
heavy-duty.
1. Light heavy-duty diesel engines are typically
rated from 70 to 170 horsepower. Vehicle
body types would encompass vans, trucks,
recreational vehicles, and some single axle
straight trucks. The gross vehicle weight
(GVW) rating of the vehicle is usually less
than 19,500 pounds [40CFR86.085-2(a)(l)].
2. Medium heavy-duty diesel engines are typi-
cally rated from 170 to 250 horsepower.
Vehicle body types include buses, tandem axle
trucks, dump trucks, etc. The GVW rating of
the vehicle is usually from 19,500 to 33,000
pounds [40CFR86.085-2(a)(2)].
3. Heavy heavy-duty diesel engines typically
exceed 250 horsepower. Vehicle body types
are typically tractor-trailer rigs and trucks and
buses used in long haul intercity operations.
The GVW rating usually exceeds 33,000
pounds [40CFR86.085-2(a)(3)]. Heavy heavy-
duty diesel engines are sleeved and designed
for multiple rebuilds.
Vehicle age and model year effects are accounted
for through the exploration of engine certification
groups as a technology group criteria: pre-1984,
1984-1987, 1988-1990, 1991-1993, 1994-1997,
1998-2003, and 2004+. These seven model year
groups represent a combination of drive train and
emission control technology integration in re-
sponse to changing emission testing standards.
1. It is possible that even within a vehicle and
engine class indicator group, there will be
specific engine or drive train elements (such
as the presence of a turbocharger, torque
converter, etc.) that will necessitate splitting
the group.
2. Manual vs automatic transmissions are han-
dled with separate indicator variables within
each technology group, due to the additional
torque converter power loss.
• Analysis of standard engine dynamometer and
second-by-second chassis dynamometer test
results, in which axle horsepower is measured or
calculated from known test parameters and
dynamometer settings, is also an important step
in the development of technology groups. Sub-
sets of vehicles within a physical class/drive
train/configuration group may behave very
differently with respect to emissions. Different
fueling strategies and engineering associated with
other control systems (engine computer software)
can result in major differences in on-road perfor-
mance from otherwise very similarly configured
vehicles.
1. Add-on control devices will likely need to be
included in the criteria as they enter the fleet
in greater proportions.
2. Once physical configurations are established,
regression tree analysis will be used to iden-
tify other variables that explain variability
across emission test results within the physical
configuration group. These factors will be
used to further subdivide the groups into their
final technology groups.
In summary, the vehicle technology groups will be
physically and statistically derived. From a physical
standpoint, the vehicles will be divided into recogniz-
able and identifiable groups based on physical con-
figuration (three major groups, plus additional sub-
groups for vehicle-trailer configuration). The evolu-
tion of drive train technologies is accounted for by
using three engine horsepower groups and seven
certification groups. These parameters account for the
major differences expected to be noted in the estima-
tion of engine load. The final subdivisions into tech-
nology group are based on statistical analysis of the
emission test databases to identify any other factors
that appear to have a significant impact on vehicle
emissions performance within each physical configu-
18
-------
Emissions Modeling Framework
ration group.
Emitter Classifications
In both the Phase I and Phase II models, vehicles will
be classified as either normal-emitters or high-
emitters. The vast majority of HDVs are normal
emitters, but a small percentage of vehicles have been
tampered with or are mal-maintained. In the Phase I
model, the EPA and other emissions testing databases
will be queried to identify those vehicles that are
high-emitters, based on emission test results for each
pollutant. As was done in the development of the
MEASURE model (Wolf et al., 1998), high-emitter
status will be defined via statistical analysis of test
results within a single certification category.
For all on-road operations, on-road vehicle activity
by vehicle class, configuration, and drive train tech-
nology will be divided into two emitter fractions.
Higher grams per brake-horsepower-hr emission
rates will be applied to the high-emitter fraction
based upon review of the literature. Hence, for each
technology group on the roadway, a high-emitter
fraction will be tracked.
In the Phase II model, the goal will be to include a
hazard-type model to predict the fraction of high-
emitters on the road as a function of accrued vehicle
mileage, vehicle age, applicable inspection programs,
and so forth. The model component is identified in
the Phase II flowchart as emitter characterization.
On-Road Traffic Volumes
A maj or problem with the heavy-duty truck emissions
modeling frameworks has been the mismatch be-
tween EPA engine certification classifications and
on-road vehicle classifications employed by the
FHWA. Mapping the two classifications to each
other is difficult because they were developed for
completely different purposes. Current guidance
suggests aggregating field observation data to total
truck volume and then disaggregating back to EPA
classifications using information from sources such
as the Polk inventory. However, Yoon (2005b)
identified a number of major problems with this
technique specifically related to the overestimation of
vehicle miles traveled (VMT) fractions for smaller
heavy-duty vehicles and the underestimation of
heavier rig VMT (Figure 6). To resolve the problems
with the current modeling framework, alternative
methodologies have been proposed.
£0.3
D Estimated HDV VMT Distribution
• 2004 MOBILE6.2 Default VMT Distribution
r. ^rl rl [I
EPA Heavy-Duty Vehicle Class
Figure 6. VMT by EPA Classification—EPA
Guidance vs Yoon Method (Yoon, 2005b).
The estimation of technology group fractions on each
roadway link begins with an estimate of total heavy-
duty diesel vehicle volumes, by EPA classification.
The goal is to develop a method that allows field
observations to be correctly mapped to EPA classes.
A key element in this process is to first identify the
on-road truck volumes using an observational classi-
fication technique designed to maximize the accuracy
of initial grouping of similar vehicles and engines. A
maj or research effort was conducted in Atlanta during
2003-2004 to develop the roadway activity estimates,
which are outlined in a recent report prepared for the
Georgia Regional Transportation Authority (Rodgers
et al., 2004). In the Atlanta effort, the researchers
developed observation categories and mapping tools
(Yoon et al., 2004a).
The visual classification scheme found to work best
for field observation is the X-Scheme, which uses
simple axle counts as the classification tool (2-axle
HDV, 3-axle HDV, and 4+ axle HDV). Given truck
19
-------
Heavy-Duty Diesel Vehicle Model
weight limitations, the axle-based classification
scheme naturally places most observed vehicles into
their medium-heavy-duty and heavy-heavy-duty
groups. Algorithms were developed to map vehicle
observations from the X-scheme to the previous
Georgia Tech classification scheme and to the FHWA
scheme. This mapping is illustrated in Figure 7,
which also indicates the applicable EPA heavy-duty
vehicle certification categories.
EPA X-Scheme
HDV2b. HDV3,
HDV4. HDV5.
HDV6, HDV7
HDVSa
X1
(2-axle)
X2
(3-axle)
r HWA
3 (HDV),
5
6.
8 (3-axle)
A
B (3-axle),
C (3-axle)
HDV8b X3 7, B (4-axle).
(>3-axle) 8 (4-axle), 9. C (4-axle),
10,11,12.13 D
Figure 7. Interactive Heavy-Duty Vehicle Mapping
Results (Yoon, 2005b).
Emissions Rate Differences
The differences of emissions rates (tons/day) between
the EPA guidance and the X-scheme scenarios were
compared (Table 2). Emissions rates with the
X-scheme scenario increased 15.7%, 34.4%, and
32.5% for VOC, NOX, and PM25 as compared to the
EPA guidance scenario. However, CO emissions rate
with the X-scheme scenario decreased 23.2% less
than that of the EPA guidance scenario.
Table 2. Emissions Rates Differences between the
X-Scheme and EPA Guidance.
Pollutant
Emission Rate Difference
voc
CO
NOX
PM9,
15.7
-23.2
34.4
32.5
The proposed methodologies use observational data,
registration databases, and survey data. Figure 8
illustrates the combined methodologies employed in
estimating heavy-duty vehicle activity. The detailed
text description of the combined system is provided
in Appendix C.
Once the EPA classifications are derived, the next
step is to develop engine classification splits and
2003 Georga Tech
HDV Database
2002 Vehicle
Inventory and Use
Survey
Mean Annual Miles
per Vehicle for
HDV3-8A Classes
Registered HDV3-
8A Classes by Fuel
Type & Age
2003 Georgia Tech
School/Urban Bus
Database
2003 Georgia Highway
Performance
Monitoring Sty stem
1993-1999
Highway Statistics
HDV VMT by Facility Type
& HDV Class
Afpfy VMT Fractions &r HDV3-SB,
, & £>&)« Bus
Figure 8. Heavy-Duty Vehicle VMT Estimation Procedure.
20
-------
Emissions Modeling Framework
vehicle configuration splits to finalize technology
groups. Engine size distributions are taken directly
from Ahanotu (1999), in which truck classes and
engine HP ratings were derived. Engine size classes
will help to further subdivide the certification groups
into those that are likely to have different emissions
performance under on-road conditions. Vehicle con-
figuration classifications are important from the
perspective of estimating road load power demand.
For example, bobtails, single unit trailers, double
trailers, flattops, and automobile carriers will all have
significantly different frontal areas and drag coeffi-
cients. Subdivisions will be developed when esti-
mated loads at high speeds are significantly different
for separate groups of vehicles within the EPA and
engine size class.
Transit Buses
In most metropolitan areas, the majority of bus miles
of travel are occurring on specific routes at specific
times of the day. The Metropolitan Atlanta Rapid
Transit Authority (MARTA) fleet is composed of
approximately 760 heavy-duty buses. Table 3 pro-
vides the distribution of the model years and their
NOX emission rates. Information is readily available
on the exact bus configuration and drive train tech-
nologies employed, and similar information is avail-
able for the Georgia Regional Transportation Author-
ity (GRTA), Clayton County, and Cobb County
fleets. Given that complete fleet and route informa-
tion can be obtained by transit agencies, the Phase I
model will employ the actual data for heavy-duty
transit buses. One interesting finding from a recent
field study in Atlanta is that the buses are not
matched to specific routes. When drivers arrive at the
staging yard, they are randomly assigned buses from
the pool. Hence, in Atlanta, there is no need to tie
specific bus technologies to specific routes.
Off-Road Activity
A significant portion of heavy-duty vehicle emissions
occur off-road, in freight yards, at truck stops, and at
pickup and delivery points. A large portion of this
activity is simple idling, usually to provide cabin
comfort (i.e., air conditioning and heating). On the
Table 3. MARTA Bus Age Distribution.
Year
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
Number
106
81
1
0
1
116
0
51
0
40
49
172
0
77
18
47
Fraction
0.1397
0.1067
0.0013
0.0000
0.0013
0.1528
0.0000
0.0672
0.0000
0.0527
0.0646
0.2266
0.0000
0.1014
0.0237
0.0619
NOX Emission
Rate
(g/bhp-hr)
4
4
4
4
5
5
5
5
5
5
5
6
10.7
10.7
10.7
10.7
other hand, emissions are usually elevated for the
first few minutes of vehicle start-up operation while
engine combustion stabilizes and emission control
systems come online, and light-duty vehicle emission
rate models are handling engine start emissions as
discrete events (as if all of the emissions associated
with the start were emitted in a puff). In the Atlanta
metropolitan area, a major review of on-road, off-
road, and non- road vehicle activity has recently been
completed. This study provides insight into engine
start and idling activity. The engine start and idle
factors for Atlanta will be built into the ArcGIS
framework using the land use layer to assign the
activity.
As discussed later in this report, engine start emis-
sions are not employed in the Phase I model due to
lack of statistically significant evidence identified in
the assessment of the EMFAC2000 model develop-
ment data. Based on the data, engine starts appear
likely to elevate emissions and may be significant
21
-------
Heavy-Duty Diesel Vehicle Model
when larger data sets are available for analysis.
However, chassis dynamometer data used in the
development of EMFAC2002 will be analyzed over
the coming year. If engine start emissions from diesel
are significant, engine start emission rates (gram per
start) will be coupled with engine start activity
estimates by parcel-level land use in Phase II model
implementation.
Idle emissions do not need to be handled through a
load estimation model. Use of grams per brake-
horsepower-hour emission rates from certification
test results will not be useful because the databases
provide only the average emission rates rather than
test component emission rates. To the extent that
second-by-second Federal Test Procedure (FTP) test
data are available, the idle emission rates can be
quantified on a grams per second basis. Idle emis-
sions can be derived from the existing second-
by-second chassis dynamometer and on-road emis-
sion testing databases. Thus, the input data required
to estimate idle emissions will be hours of vehicle
idle per technology group by land use activity.
A potentially more significant and difficult modeling
problem is the estimation of emissions associated
with freight yard creep activity. The high inertial
loads associated with repetitive acceleration as
vehicles move through queues will likely need to be
quantified through new field studies. The standard
load-based modeling approach can be applied once
hours or miles of creep activity are compiled and
speed/acceleration distributions are defined for this
activity.
22
-------
Emissions Modeling Framework
Freight and Passenger Loads
The load-based modal model requires estimates of
vehicle weight for each technology group. The
combined weight of the vehicle and the trailer signifi-
cantly affects power demand through the grade load
and road friction terms.
Truck Loads
The combined weight of vehicle and trailer is em-
ployed in road load calculations (road friction and
grade load). Hence, characterization of vehicle capac-
ity and utilization factors (i.e., actual on-road vehicle
weights) is required input to the model. However,
most emissions-related HD V classifications are based
on gross vehicle weight rating (GVWR) as a result of
the availability of accessing this data from vehicle
registration data bases. These ratings are based on the
maximum weight that a vehicle can carry relative to
the horsepower capacity and safety considerations of
the vehicle. Many HDVs are classified with the same
weight rating despite having different on-road activ-
ity, weight, and horsepower characteristics. For
example, a vehicle with a GVWR of 80,000 pounds
can be a 5-axle, long-haul, tractor-trailer combination
with a 600 horsepower engine used for dry goods
shipment or can be a 3-axle, local, and single-unit
vehicle with a 350 horsepower engine used for
hauling loose materials. The on-road weight distribu-
tion of the 5-axle vehicle would range from an empty
weight of 35,000 pounds up to the legal maximum of
80,000 pounds. The on-road weight distribution of
3-axle vehicles generally ranges from an empty
weight of 20,000 pounds up to the legal maximum of
54,000 pounds. So, heavy-duty vehicle classification
methods simply based on gross vehicle weight rating
do not match on-road vehicle activity categories.
Between 1997 and 1999, Ahanotu (1999) undertook
a series of weigh-in-motion studies and weigh station
interviews to examine the spatial and temporal
distribution of vehicle loads in the Atlanta region.
These study results serve as the basis for integrating
truck loads into the heavy-duty load-based modeling
framework. Ahanotu (1999) developed a modified
truck classification based on axle-trailer configura-
tions which allows for the development of weight and
horsepower distributions. The vehicle classification
scheme—a four-class system based upon axle-trailer
configurations—will use the FHWA classification as
a base for all ground counts and field observations:
• A: 2-axle, single unit trucks (Class 5 trucks)
• B: 3 or more axle, single units trucks (Class 6-7
trucks)
• C: 3 or 4 axle, truck and single trailer combina-
tions (Class 8 trucks)
• D: 5 or more axle single trailer trucks, plus
double trailer trucks (Class 9-13 trucks)
The four observational classifications are composed
of FHWA vehicle classes that were aggregated based
on three criteria: (1) range of gross weights across
truck classes, (2) engine horsepower differences
across truck classes, and (3) ease of classifying
during field observations (Rodgers et al., 2004).
Ahanotu (1999) developed this scheme with relation-
ships between engine horsepower and gross vehicle
weight in mind. From the survey data collected by
Georgia weight enforcement stations, Ahanotu found
that the Class 9-13 truck classifications averaged
significantly higher horsepower ratings (about 370)
relative to the Class 8, Class 6-7, and Class 5 truck
classes, which had average ratings of 293, 279, and
188 hp respectively. Figures 9 and 10 illustrate the
horsepower distributions and average values by
23
-------
Heavy-Duty Diesel Vehicle Model
FH\\
engir
engir
45-
>~ 30 -
^A truck class. Figure 11 illustrates the shift in 0.400 -
ie horsepower ratings over time, as newer
les have become larger. £• , „ .
IS
O A 1 r A -
— * — Class 9- 13 o-
* A -j f) A _
— m— A LLIUU
-—•— Class 6-7 "7 \~ 0.050 -
••*«••• CBSS 5 / \ U.UUO -1
/ \
.•"'••- /" \
"•x / \
• -^•^— — »^_ \
^ m^*-^3?'^---'t "! ^^f^-"- . »,..___, Finiirp 11
_ —•—pre-1996
fl — ,1, — 1 998 and newer
i 1
/ jff\
/A\ /\x
r i/ \\ 4 \ \
t>| f^, ^i f^.» fsj r^- CM N~
CM (M ell to TJ- -* LO tf)
Upper Boundary Horsepower Bin
. Fnninp Horspnowpr Di.cstrihiition.c! hv
100-149 150-199 200-249
Horsepower Ranges
Figure 9. Horsepower Distributions by Class for
Preliminary Surveys.
450499 Model Year Grouping.
Figure 10. Mean Horsepower by FHWA Truck
Class.
Ahanotu's analyses indicated that a 4-class HDV
classification scheme would be suitable for describ-
ing both horsepower and weight distributions. This
classification system allows data from numerous
truck data sources (i.e., roadside surveys, vehicle
counts, emission rate estimates, commercial vehicle
surveys, etc.) to be easily incorporated into emissions
models. However, Yoon et al. (2004a) later deter-
mined that a revised system (X-scheme, or axle
scheme) would potentially improve on-road classifi-
cation recognition. This is because the previous
system was more difficult to map back to the EPA
heavy-duty vehicle boundaries (based upon GVWR).
The X-scheme better bridges the FHWA truck and
EPA HDV classification system. Table 4 shows the
mapping between the classification schemes.
Table 4. X-Scheme—HDV Reclassification Map
Among HDV and Truck Classification Schemes.
„. X EPA Classes
Classes
XI HDV2b,HDV3,
HDV4, HDV5,
HDV6, HDV7
X2 HDVSa
FHWA
Classes
3 (HDV),
5
6,
Ahanotu s . ,
„, Axles
Classes
A 2
B (3-axle), 3
X3
8 (3-axle) C (3-axle)
HDVSb 7, B (4-axle),
8 (4-axle), C (4-axle),
9,10,11, D
12, 13
>3
The XI class includes pick-ups, vans, and delivery
trucks. The X2 class includes dump trucks and
articulated delivery trucks having three axles. The
X3 class includes all more than three-axle, articulated
or single HDVs. With the X-scheme, FHWA truck
classes can be more accurately mapped into EPA
HDV classes, or vice versa.
In developing the X-scheme, corresponding engine
horsepower data were not available. Hence, it is not
24
-------
Emissions Modeling Framework
clear at this time how to best map the horsepower
distributions back to the X-scheme vehicle classes.
For the time being, it will be assumed that the horse-
power distributions determined through Ahanotu's
research can be allocated proportionally to the
X-scheme. That is, if 40% of one Ahanotu class is
now allocated to class X3 and 60% of another class
is mapped to X3, the X3 engine horsepower distribu-
tion will be a weighted sum of the Ahanotu distribu-
tions. The noted effect of engine age on horsepower
distributions will be similarly integrated. As com-
bined engine horsepower, model year, and classifica-
tion (X-class) data become available in future, up-
dated horsepower distributions will be integrated into
the model.
Ahanotu (1999) also examined the spatial and tempo-
ral distribution of engine horsepower and GVWR for
the Atlanta metropolitan area. These analyses can be
directly integrated into the model, given the spatial
and temporal nature of the modeling framework.
Figure 12 illustrates the load distribution of trucks
operating during the early afternoon period, with the
weight distribution being a combination of empty,
full, and partially loaded trucks. Ahanotu's research
also provides temporal effects, with different distribu
tions applied in the morning, noon, and afternoon
periods (more fully-loaded trucks are operating in the
morning peak). Figure 13 shows the differences in
these distributions by day of week. Ahanotu's on-
>• fifi -
(j W
= M
e
1 kip = 4448 newtons \
A
A
t\
s \
r \ / \
Y7
/ V
i
^ * r
v v i
\
.*.vM^ V,^..
Weight (kips)
03
CT
£
Figure 12. Class 9-13 Weight Distribution, Monroe
County Weigh Station (4/12/98, 12-2PM).
Weight Bins (1000 Ibs)
Figure 13. Midday Time Period Weight Distri-
butions, Class 9-13 Trucks.
road findings for vehicle load will be integrated into
the model directly so that the road load power de-
mand can be established for each class using the data
in the load matrices (by hour of day and day of week
for each class).
Considering that all of the engine horsepower, vehi-
cle class, and weight distribution data were collected
in 1998, it would be advisable to perform this re-
search effort again in the future to ensure that rela-
tionships have not changed. However, in the absence
of updated data, the research team will continue to
use Ahanotu's results.
Transit Passenger Loads
Initial estimates for Atlanta indicate that passenger
loading can have a 10% to 15% impact on load-
predicted emissions. Transit passenger loads can
either be provided to the model via input from the
4-step travel demand model (in which each transit
route is its own link) or be manually entered based
upon field studies. In Atlanta, MARTA has con-
ducted detailed passenger loading studies and can
provide boarding and alighting data for each route. In
the absence of such agency data, it is a relatively
simple matter to conduct such studies in an urban
area.
25
-------
Heavy-Duty Diesel Vehicle Model
26
-------
Emissions Modeling Framework
On-Road Operating Characteristics
The heart of the implementation approach for the
modal model is the application of speed/acceleration
frequency distributions (SAFDs), which describe the
fraction of vehicle activity that occurs under specific
speed and acceleration conditions. The matrix is
composed of cells that are binned by speed (0 to 2.5
mph, 2.5 to 5 mph, and then 5 to 100 mph in 5 mph
increments) and acceleration (0 to 10 mph/s in 0.5
mph/s bins). In the model implementation described
here, each second of vehicle activity is assigned to a
single speed acceleration matrix cell. Idle activity is
defined as vehicle activity falling into four cells,
bounded by -0.25 to 0.25 mph and -0.5 to 0.5 mph/s.
The definition of cruise activity is somewhat arbi-
trary, but a reasonable range for cruise might be
non-idle activity at any speed falling into an accelera-
tion range of-1.0 and 1.0 mph/s. Figure 14 illustrates
a speed/acceleration plot for an interstate ramp
operating at level of service D (peak afternoon
commute) in the MEASURE modeling system based
on data collected for the Atlanta Ramp metering
Study (Guensler et al., 2001).
5 Velocity (mph)
Acceleration (mph/s)
Figure 14. Speed/Acceleration Profile, Interstate
Ramp, LOS D (Bachman, 1997).
The goal is to represent the vehicle activity on any
link at any given time as total traffic volume (flow
per hour) under a specific set of speed and accelera-
tion conditions. That is, a specific SAFD should
apply to any roadway for each hour of the day and
day of the week. The data necessary to create the
SAFDs to represent on-road operating conditions can
be derived from a variety of on-road monitoring
methods and techniques, the most common sources
being stationary laser gun studies or instrumented
vehicle studies (chase car or instrumented fleets).
HDV acceleration rates are significantly lower than
LDVs (an interactive function with vehicle load). For
example, the maximum acceleration rate from 0 to 50
mph for a 100 Ib/hp tractor-trailer is 1.0 mph/s2, with
decreasing acceleration rate as the initial speed is
higher, compared to a compact LDV with an acceler-
ation rate of 5.3 mph/s2 (Grant, 1998). Thus, different
speed/acceleration profiles should be used for light-
duty and heavy-duty vehicles, especially on freeways.
For freeways and ramps, the Phase I model will
employ basic speed/acceleration distributions devel-
oped for the MEASURE model from Atlanta metro-
politan area data collected by Grant (1998). Grant's
research focused on the development of statistical
relationships between power surrogates (positive
kinetic energy and wind load surrogates) using
second-by-second vehicle data collected by the
extensive laser gun study. In his research, power
surrogates were determined as a function of road
type, location (central business district, suburban, or
rural), design speed, number of lanes, lane width,
shoulder width, grade, congestion levels, and pres-
ence of trucks (on grades). Plus, Grant (1998) ob-
served that the presence of trucks on uphill freeway
27
-------
Heavy-Duty Diesel Vehicle Model
grades affects not only the truck speed and accelera-
tion conditions, but the speed and acceleration condi-
tions of the LDVs operating on the system as well.
Thus, the speed/acceleration matrices employed in
link-by-link calculations must also incorporate the
effect of grade. Because the focus is now to generate
speed and acceleration profiles and calculate power
demand directly, rather than estimate power demand
surrogates from the data and apply them to MEA-
SURE modal emission rates, the second-by- second
data will be de-archived and reanalyzed to develop
more refined freeway speed and acceleration profiles
for the Phase II model.
For arterials and local roads, the research team plans
to use the Commute Atlanta instrumented vehicle
data to represent on-road operating conditions for
HDVs. The Commute Atlanta study collects approxi-
mately two million vehicle-seconds of on-road
operating data every day from a fleet of 460 instru-
mented LDVs. In addition, two transit buses operat-
ing on MARTA routes have also been instrumented.
The trip data collector consists of a global positioning
systems (GPS) receiver, a wireless communication
device, data storage, and on-board diagnostic (OBD)
systems. As a new, emerging vehicle speed data
collection tool in transportation research field, GPS
receivers provide highly accurate speed data calcu-
lated with the Doppler shift theory. In studies of
vehicle speed accuracy using GPS receivers, vehicle
speed from GPS receivers is as accurate as that
obtained from a conventional distance measuring
instruments or travel time data acquisition systems.
Although the LDV speeds and accelerations are
different than those of heavy-duty trucks, until
enhanced truck operating profile data are made
available, the team believes that these profiles should
be used.
The research team recently developed transit bus
speed-acceleration matrices using speed data ob-
tained with a Commute Atlanta trip data collector
(Yoon et al., 2005a). Given the limited number of
routes driven by transit buses in major metropolitan
areas, the instrumented vehicle approach is recom-
mended by the research team. The application of
these speed acceleration profiles is described in the
case study chapter of this report.
28
-------
Emissions Modeling Framework
Engine Power Functions
Internal combustion engines translate linear piston
work (force through a distance) to a crankshaft,
rotating the crankshaft and creating engine output
torque (work performed in angular rotation). The
crankshaft rotation speed (engine speed in revolutions
per minute) is a function of engine combustion and
physical design parameters (mean effective cylinder
pressure, stroke length, connecting rod angle, etc.).
The torque available at the crankshaft (engine output
shaft) is less than the torque generated by the pistons
because there are torque losses inside the engine
associated with operating a variety of internal engine
components. Torque is transferred from the engine
output shaft to the drive shaft via the transmission
(sometimes through a torque-converter, or fluid
coupling) and through a series of gears that allow the
drive shaft to rotate at different speeds relative to
engine crankshaft speed. The drive shaft rotation is
then transferred to the drive axle via the rear differen-
tial. The ring and pinion gears in the rear differential
translate the rotation of the drive shaft by 90 degrees;
from the drive shaft running along the vehicle to the
drive axle that runs across the vehicle. Torque avail-
able at the drive axle is then delivered directly to the
drive wheels, which generates the tractive force used
to overcome road friction, wind resistance, road grade
(gravity), and other resistive forces, allowing the
vehicle to accelerate on the roadway. Figure 15
illustrates the primary components of concern.
Vehicle performance depends on how much of the
available engine torque can be transmitted to the
Tr»nimi««ion.
manual or automatic,
has gearsets that
match enotne speed
to desired road
speed
Engirt*
provides the power
(torque x speed) to
propel the vehicle
via the drivetraln
Axleihufl
turning Inside each
rear axle housing lube
transmits power from
the differential to Mis,
rear wheels
Ml houiing
contains the clutch
for e manual
transmission or
the torque converter
tor tn automatic
transmission
DrivMhaft
passes power from
the transmission to
Differential
turns power flow
90 degrees and allows
the differentia! housing, one wheel to rotate
Ujoints allow it to (aster than the other
ride up and down on curves or when
with the rear «xto traction differs
Figure 15. Primary Elements in the Drive Train (Gillespie, 1992).
29
-------
Heavy-Duty Diesel Vehicle Model
wheels and used to overcome the resistive forces
acting against the vehicle. Torque losses arise in the
conversion of engine torque to wheel torque. Work
must be performed to overcome resistive and inertial
forces within the drive train system. Plus, engines are
often called upon to provide power to operate acces-
sories (e.g., air conditioning or refrigeration compres-
sors).
Hence, engine load is composed of observable road
load components (to perform the work necessary to
overcome external forces such as wind resistance and
road friction) as well as for internal components
associated with overcoming drive train friction and
component inertia and for running accessories.
Power is a measure of how quickly work is per-
formed. Tire rotation (i.e., wheel work) is equal to the
force delivered by the wheel torque (torque divided
by wheel radius) times the distance traveled by the
tire (2ft times the radius), or wheel torque times the
angular distance per rotation (2jr). The power is then
the work times the rotational speed (in revolutions
per second), where the rotational speed of the tire is
a function of engine speed and gear ratios. The
formula for axle power is
Ihp
550
ft •
sec
(1)
each rotation of the engine crank shaft provides a
different number of rotations of the drive shaft and
then of the drive axle (and therefore the drive
wheels). Transmission and differential gear ratios
allow the wheels to rotate at lower rotational speeds
than the engine (which may be operating at more than
4000 rpm). The gear ratio at the differential (G^), or
final drive unit, can range from as low as 3.5:1 to as
high as 5.4:1. Higher differential ratios provide
greater torque for towing, and lower ratios provide
better on-road fuel economy at high speeds. Trans-
mission ratios range from around 2.7:1 (first gear)
down to 0.75 (overdrive). Larger heavy-duty trucks
tend to operative with five or six gears and with
larger ratios in first gear to provide additional torque
multiplication to accelerate from a standing stop. In
first gear, rotational speed is lower, but torque output
is higher. Each wheel rotation results in a distance
traveled that is a function of wheel radius. Vehicle
velocity is
(2)
60 sec)
where Fis vehicle speed in feet per second, and
r is the radius of the drive wheel in feet.
The relationship between engine speed, gearing, and
vehicle velocity can be used to translate the axle
power equation to
where PA is axle-horsepower (ahp) available for
tractive work,
TA is torque available at the drive axle in
foot-pounds force,
NE is the engine speed in revolutions per
minute (rpm),
G,is gear ratio at the engine transmission, and
Gd is gear ratio in the final drive (differen-
tial).
Vehicle velocity is determined directly by engine
speed, gearing, and drive wheel radius. Depending
upon the operating gear ratio and differential ratio,
VxT,
P = —
A rx550
(3)
At any given instant, the axle horsepower available
for tractive work can be quantified through observa-
tion of a vehicle's activity. Vehicle velocity can be
observed directly. Given that axle torque provides the
vehicle motive force at the pavement (torque divided
by wheel radius), axle power can be re-written as
550
(4)
30
-------
Emissions Modeling Framework
where Fm is the motive force available at the pave-
ment in pounds force.
The motive force delivered at the pavement over-
comes the variety of resistance forces (road friction,
wind resistance, road grade, etc.). These forces can
also be quantified if specific vehicle and environmen-
tal parameters are known. Any remaining motive
force that is in excess of the resistance forces pro-
vides vehicle acceleration. Because the vehicle accel-
eration rate can also be observed, road load horse-
power can be modeled as a function of vehicle and
operating environment characteristics.
Engine power is equal to the accessory power loss,
drive train power loss, and available axle power
(5)
where PE is engine brake-horsepower,
PDT is brake-horsepower loss in the drive
train, and
Pa is brake-horsepower loss from accessory
operations.
Overall brake-horsepower demand on the engine can
be estimated when power losses associated with the
drive train and accessory operation can be quantified
and when axle horsepower demand can be predicted
as a function of vehicle characteristics and observed
on-road operating parameters. As discussed in the
model overview section, once the brake-horsepower
demands are quantified, engine dynamometer emis-
sion rate relationships (gram per brake-horsepower-
hour) can be used directly in emissions modeling.
Larger and larger heavy-duty vehicle emission testing
data sets are currently being developed on chassis
dynamometers and through on-road testing in which
axle horsepower loads (axle horsepower per second)
are being measured concurrently with emission rates
(grams per second). Chassis dynamometers measure
the change in roller drum speed (drum acceleration)
produced by excess tractive force delivered by the
heavy-duty drive wheels. Axle torque can be calcu-
lated, given the rotational inertia of the roller drum.
Axle torque and drum speed provide axle horse-
power. When such data sets are sufficiently robust,
axle-load emission rates (grams per axle-horsepower-
hour) can be employed directly without the need for
estimating the drive train and accessory power losses.
This will require strict accounting within vehicle
technology groups to ensure that all vehicles within
a technology group are substantially similar with
respect to both drive train power loss relationships
and parasitic accessory loads, but modeling should
improve as a result.
Accessory Power Loss
Engines dynamometer tests measure engine output
torque and engine speed (revolutions per minute), and
engine horsepower output can be calculated as
TExNE
5252
(6)
However, the gross torque and power output reported
from these tests represent the operation of the engine
with only its internal engine accessories in place.
Complete intake and exhaust systems add additional
resistance to working gases. The maximum on-road
engine power output for an engine equipped with
complete intake, exhaust, and cooling systems can be
14% less than the maximum gross power output
(Heywood, 1988).
Air conditioning compressor unit physical design and
operating pressures determines the engine torque loss
(and associated power loss) from cabin comfort
operations. In LDVs, as much as 15 peak horsepower
can be lost from running an air conditioner. More-
over, air conditioning compressors often operate
intermittently, compressing and throttling fluids as
needed, meaning that the power demand on the
engine for air conditioning can vary.
About 15% to 20% of the available engine horse-
power can be lost in the process of overcoming
internal backpressures, providing cabin comfort, and
31
-------
Heavy-Duty Diesel Vehicle Model
operating other engine/cabin peripherals. To predict
engine load, any observations of road load horse-
power would need to be adjusted upwards to account
for these accessory power losses. The Phase I model
employs a very simplistic approach to account for
accessory losses, which are equivalent to the sum of
SAE default power requirement from individual
accessories (SAE, 2004).
Engine power takeoffs can also deliver power neces-
sary to operate refrigeration units, alternators, genera-
tors, or other major devices. Unfortunately, little
information is currently available in the literature
regarding power consumption associated with power
takeoffs. Additional power losses associated with
other accessory loads will not be implemented in the
Phase I model. The Phase II model provides the
capabilities to implement accessory power loss algo-
rithms as a function of simulated air conditioning
draw (as a function of temperature and humidity for
each vehicle technology) and will be able allocate
large accessory loads to specific heavy-duty vehicle
configurations (such as refrigeration trucks). This is
accomplished by running each technology group
through the power loss calculation, weighted by the
fraction of vehicles to which the algorithm applies.
The ongoing literature review, future truck stop
inspections to identify accessory use relationships,
and data becoming available from agencies regarding
air conditioning usage and power loss will help to
quantify these relationships.
As emission testing provides improved grams per
axle-horsepower-hour emission rate relationships,
vehicle class-configurations combinations that have
large accessory loads can be modeled as their own
unique technology groups. This would allow field test
axle-horsepower emission rate data to be used di-
rectly without further quantification of these acces-
sory loads because the loads are already inherent in
the emission rate relationship, essentially in the
intercept term.
Drive Train Power Loss
After engine horsepower at the output shaft has been
reduced by power losses associated with fluid pres-
sures, air conditioning operation, and other accessory
loads, there is still an additional and significant drop
in available power from the engine before reaching
the wheels. Power is required to overcome (1) me-
chanical friction in the transmission and differential
and internal working resistance in hydraulic cou-
plings, and (2) friction of the vehicle weight on axle
bearings. The combined effect of these components
will be described as drive train efficiency. The more
difficult and more significant component of power
loss in the drive train is associated with the inertial
resistance of rotating drive train components and
their inertial resistance to acceleration.
Mechanical Friction
Friction is inherent in the fluid and mechanical
couplings in the drive train (clutch, transmission, and
differential). Power demand to overcome these
friction forces is a function of the specific technolo-
gies employed. Laboratory testing can readily quan-
tify the power demand to overcome these forces,
which are functions of drive train design and material
composition. Although each engine and vehicle
manufacturer undertakes specific testing to quantify
these forces (and to develop system improvements
designed to reduce them), little information on
mechanical drive train loss by vehicle technology is
readily available. Most sources show the combined
power loss through the drive train system, without
isolating the mechanical friction effects.
Torque Converter Losses
Automatic transmissions employ torque converters (a
fluid coupling between the engine output shaft and
transmission) to transfer rotation from the engine
output shaft to the transmission, and power loss in
them is a function of input and output rotation speeds.
Large differences (i.e., when starting from a standing
stop and shifting gears) lead to large power loss.
Higher rotational speeds can yield around a 3% loss
in power at high speed (best performance conditions).
The model will assume that a drive train equipped
with torque converters will experience an additional
5% drop in available engine horsepower.
32
-------
Emissions Modeling Framework
Axle Bearing Friction
The weight of a vehicle must be supported by the
tires at the roadway surface. Axle bearings transfer
the weight of the vehicle and cargo to the axles, and
therefore to the tires, while still allowing the axles to
spin freely within the bearing with a minimum of
frictional resistance. Due to current bearing technol-
ogy, this frictional resistance is minimal. Thus, axle
friction will not be included in the modeling frame-
work.
Inertia! Losses
The engine, transmission, drive shaft, axles, and
wheels are all in rotation. The rotational speed of
each component depends on the transmission gear
ratio, the final drive ratio, and the location of the
component in the drive train (i.e., the total gear ratio
between each component and the wheels). The
rotational moment of inertia of components in the
drive train constitutes a resistance to change in
motion. The torque delivered by each rotating com-
ponent to the next component in the power chain
(engine to clutch or torque converter, clutch or torque
converter to transmission, transmission to drive shaft,
drive shaft to axle, axle to wheel) is reduced by the
amount necessary to increase angular rotation of the
spinning mass during vehicle acceleration. Work has
to be performed to accelerate these rotating compo-
nents. Given the torque loss at each component
(Gillespie, 1992), the reduction in motive force
available at the wheels due to inertial losses along the
drive train can be modeled as
ax
(G2d
(7)
where a is the acceleration of the vehicle in the
direction of motion in feet per second
squared,
Fj is force required to overcome inertial
resistance in pounds force,
Iw is the rotational moment of inertia of the
wheels and axles in foot-pounds (force)-
second squared,
ID is the rotational moment of inertia of the
drive shaft in foot-pounds (force)-second
squared,
IT is the rotational moment of inertia of the
transmission in foot-pounds (force)-second
squared, and
IE is the rotational moment of inertia of the
engine in foot-pounds (force)-second
squared.
Inertial resistance is a function of speed, gearing,
acceleration, and the specific drive train technologies
that affect the mass moment of inertia components of
the equation. This force component can be employed
directly to estimate power loss from inertial resis-
tance
VxF,
p- -
An alternative, but not recommended, approach to
modeling inertial power loss is the use of "effective
vehicle mass" in all load calculations. This approach
indirectly accounts for the inertial load impact by
artificially increasing the vehicle mass. By increasing
the effective vehicle mass of the vehicle, the available
power for acceleration is decreased, and the same
effective decrease in power can be incorporated into
the model. However, implementation of the indirect
approach requires reasonable knowledge of the
inertial load relationship for the engine and drive
train employed.
To account for effective vehicle mass, the effective
moment of inertia is first calculated as
IEff = [lw + (Gl x ID) + (G? x Gj) x (IE + IT)]
(9)
where IEffis the effective moment of inertia in foot-
pounds (force)- second squared.
The effective moment of inertia for the drive train is
a function of the moments of inertia of the drive train
components as well as the on-road vehicle operating
conditions (transmission gear ratio). The gear ratio of
33
-------
Heavy-Duty Diesel Vehicle Model
first gear is much higher than fifth gear; moreover,
first gear is physically larger and has higher compo-
nent moment of inertia. Given that the inertial terms
employ gear ratio squared, the effective inertia in first
gear can be a factor of 40 higher than the inertia in
overdrive.
The effective inertia can be translated into an effec-
tive weight
(10)
where MEffis the effective mass in pounds (mass),
WEffis the effective weight in pounds (force),
and
g is the acceleration of gravity (32.2 ft/s2).
This effective weight for any given operating condi-
tion can be added to the vehicle weight in developing
road load power demand estimates. The effective
vehicle weight increase applies only to the drive
vehicle and not to towed trailers. The impact of gear
selection on effective vehicle weight is illustrated in
TableS.
Table 5: Typical Percent Increase in Effective Vehi-
cle Weight (Excluding Trailer).3
Operating Gear
v eiutie •
Small Car
Large Car
Truckb
High
11
9
9
Second
20
14
12
First
50
30
60
Low
140
150
"Source: Gillespie, 1992
b Truck classification not specified.
Gillespie (1992) provides an example calculation for
a passenger car illustrating an effect of effective
weight contribution due to inertial resistance in first
gear of around 880 pounds (or 35% of the vehicle's
stationary weight), yet the effective weight contribu-
tion in fifth gear is only around 300 pounds (or 12%
of the vehicle's stationary weight). A heavy-duty
truck drive train is significantly more massive than its
light-duty counterpart. Nevertheless, the net effect of
drive train inertial mass losses, relative to the weight
of the vehicle, when operating in higher gears on
freeways, may not be significant enough to include in
the model. However, recent studies have shown very
high truck emission rates (grams per second) in
"creep mode" stop and start driving in ports and rail
yards. This may indicate that the inertial loads associ-
ated with accelerating drive train rotation in low gear
operations may be the most significant factor contrib-
uting to emissions from mobile sources in freight
transfer yards.
It is important to keep in mind that the effective
vehicle mass approach relies on knowledge or mea-
surement of the engine and drive train inertial resis-
tance. Thus, it makes much more sense to model the
inertial resistance directly than it does to use an
effective mass surrogate.
Driver Behavior
The driver behavior element is planned as a potential
enhancement to the inertial load component of the
load calculations. Drive train inertial loss depends on
transmission gear in which the vehicle is operating
for a given speed and acceleration condition. The
base approach is to model gear selection for each cell
as a probability function. That is, for a given speed
and acceleration bin, the vehicle is 80% likely to be
in second gear and 20% likely to be operating in 3rd
gear. Such estimates will be based solely on empirical
evidence from on-road studies. The driver behavior
enhancement would predict the likelihood of operat-
ing in a specific gear (and therefore at a specific
rotations per minute) as a function of driver demo-
graphic characteristics and experience level. This
approach cannot be implemented without a substan-
tial data collection effort. However, once in place, the
module would allow policymakers to model the
potential effects of driver training on emissions and
fuel consumption.
34
-------
Emissions Modeling Framework
Drive Train Power Loss Modeling Ap-
proach
Each drive train is developed and configured as a
system. The transmission and differential systems are
'tuned' to run with the specific engine and tire radius.
Engines achieve maximum torque at different engine
speeds, but the gearing systems and tires determine
the on-road performance of these engines. A single
drive train system may be used on many vehicle
configurations, and the weight of the various vehicle
configurations will significantly affect the vehicle
performance. Combined power loss due to the trans-
mission and differential systems can range from 10%
to 15%. Older vehicles lose significantly greater
power through the gearing system due to older design
parameters used in gear tooth profiles. The losses are
a function of a wide variety of physical drive train
characteristics (transmission and differential types,
component mass, etc.) and on-road operating condi-
tions.
Drive Train Efficiency
Given any set of specific coupling, transmission, and
differential technologies, the power losses associated
with the drive train mechanical friction and axle
bearing friction can be determined through laboratory
testing. The total efficiency of power transmission
from the engine to the road (t|tot) can be determined
by experiment, provided that parameters necessary to
separate rotational moment of inertia losses from
drive train mechanical losses are collected or known.
In the Phase I model, drive train frictional losses will
be set at 4%, with the remaining average efficiency
loss from the drive train will be assumed to arise
from rotational inertia losses.
As with the accessory load module, the Phase II
model provides the capabilities to implement differ-
ent drive train loss modules as a function of drive
train configuration (i.e., based upon laboratory test
results as data become available), vehicle configura-
tion, and vehicle weight. This module will likely
become increasingly important in the future because
the same resistive forces that increase engine load
also increase fuel consumption. Technologies de-
signed to reduce losses in the drive train have signifi-
cant paybacks in fuel savings, and some will pay for
themselves over the lifetime operation of the vehicle.
This means that the technology makeup of the fleet is
very likely to continue to evolve.
Rotational Moment of Inertia Losses
In the Heavy-Duty Diesel Vehicle Modal Modeling
Framework, drive train rotational losses are calcu-
lated as a direct power loss rather than using the
method of modifying vehicle mass in subsequent road
load power equations. This method was selected
because the inertial power demand is a direct function
of drive train component design (component weights
and rotational speeds) and independent of the factors
that affect road load power demand, with the excep-
tion of transmission gear selection. In either method,
gear selection would need to be addressed, so the
direct power loss function was deemed more practi-
cal.
Drivers can achieve a given instantaneous speed and
acceleration condition in more than one gear, using a
higher gear ratio (e.g., 2nd gear) and lower throttle
position, or lower gear ratio (e.g., 3rd gear) and higher
throttle position. However, for any given instanta-
neous speed and acceleration condition, the driver is
much more likely to be in one gear than another. This
is because the next second of operation depends on
the current operation. Selection of the most favorable
gear ensures that the vehicle will continue to cruise,
accelerate, or decelerate at close to the same rate.
In the Phase I model, the drive train losses will be
determined for the fleet. The team will assemble the
specifications and performance tests of approximately
twenty heavy-duty drive train configurations and will
calculate the inertial power losses as a function of
various speed/acceleration on-road operating condi-
tions. The researchers will either assume gear selec-
tion, or use gear selection information from on-road
tests. The average power loss results for each speed/
acceleration matrix cell will be calculated for various
vehicle configurations. The power loss associated
with drive train moment of inertia will then be
35
-------
Heavy-Duty Diesel Vehicle Model
available for each vehicle configuration and speed/
acceleration condition and can be added to the road
load power demand and accessory loss to estimate
total engine power demand.
In the Phase II model, provided that the inertial
power loss is significant enough to warrant complete
integration rather than parameterization, the drive
train power loss function will employ
• Gear selection probability matrices for each drive
train technology class (lookup tables that provide
gear ratio probability by speed/acceleration
matrix cell),
• Gear and final drive ratio lookup tables for each
drive train technology class,
• Gear and final drive moment of inertia lookup
tables for each drive train technology class, and
• Power loss matrices (lookup tables specifying
power loss as a function of speed and accelera-
tion matrix cell, given the gear selection, gear
ratios, and moments of inertia).
Axle Horsepower Relationships
To the extent that emission testing can provide
improved grams per axle-horsepower-hour emission
rate relationships, combinations of drive train tech-
nology and vehicle configurations can be further
refined into technology groups. Axle-horsepower
emission rate data from field testing could be used
directly for each technology group. However, average
axle-horsepower emission rates cannot be employed.
Specific axle-horsepower emission rates will need to
be defined by speed/acceleration matrix cell, to
ensure that inertial load power losses emission rates
are accounted for. The effects of transmission and
drive train resistance would become parameterized
within the emission rate relationship as part of the
intercept term for the technology group.
System Monitors
To the extent that engine onboard diagnostic data
become available from instrumented truck and bus
studies, researchers will be able to factor in sensor
readings in the hazard component failure models. For
example, when onboard systems identify a drop in
rpm for given conditions, the change may indicate a
maintenance problem, increasing the likelihood of
component failure. Once in place, such modules
could be used to evaluate the potential benefits of
instrumented vehicle automated inspection and
maintenance requirements. This will not be included
in the Phase I or Phase II models, since such data are
not yet available.
36
-------
Emissions Modeling Framework
Road Load Power Functions
For a vehicle to remain in motion, the tractive force
delivered at the drive wheels must overcome resis-
tance forces, including
• Tire rolling resistance forces (FR), to overcome
losses associated with tire/road friction and tire
deformation,
• Gravitational force (Fw) associated with vehicle
weight when operating on a grade,
• Aerodynamic drag force (FD~) associated with air
resistance to vehicle motion, and
• Curve resistance forces (Fc~) to overcome the
additional frictional force associated with turning
the vehicle.
Providing the exact amount of tractive force to
balance the resistive forces will maintain the vehicle
in motion at its existing speed. Any extra tractive
force delivered to the wheels will accelerate the
vehicle according to Newton's 2nd law
m x a = FT - FR- Fw - FD- Fc
W
— x a = FT - FD - F,,f - Fn - F^
(11)
where m is the vehicle mass in pounds (mass), and
FT is the tractive force available at the
wheels in pounds force.
The road load prediction component is the heart of
the modal model. Engine technologies and on-road
operating conditions affect the ability of the vehicle
to translate available horsepower into speed and
acceleration performance. Given that the on-road
speed/acceleration patterns can be observed (or
empirically modeled), the modal modeling approach
works backwards from observed speed and accelera-
tion to estimate the tractive force (and power) that
was available at the wheels to meet the observed
conditions. Then, working backwards from tractive
force, the model accounts for additional power losses
that occurred between the engine and the wheels to
predict the total brake-horsepower output of the
engine. Each force component that reduces available
wheel torque and tractive force is discussed in turn.
Rolling Resistance Force (FR)
Rolling resistance force (FR) is the sum of the force
required to overcome the combined friction resistance
at the tires. Tires deform at their contact point with
the ground as they roll along the roadway surface.
Rolling resistance is caused by contact friction, the
tires' resistance to deformation, aerodynamic drag at
the tire, and so forth. Deformation and friction
resistance are functions of tire size and type, pave-
ment type, temperature, vehicle weight, and vehicle
speed. The force required to overcome the total
rolling resistance can be expressed with the coeffi-
cient of rolling resistance (Cr), total vehicle weight
(W), and road grade
Wx cos(#)
where #is the roadway grade angle
(12)
The calculations can be performed at each tire, using
the dynamic weight component attributed each tire
(calculated with using the location of the center of
mass and moments relative to the wheel contact
points). However, given the linear relationship
between weight and resistance and given that neither
the coefficient of rolling resistance nor grade varies
37
-------
Heavy-Duty Diesel Vehicle Model
from tire to tire, there is no need to estimate for each
tire. The net coefficient of rolling resistance can be
estimated as a function of vehicle speed and the road
surface coefficient by using the SAE equations for
various tire types (Tables 6 and 7).
speed(mi/h)
(13)
where bl and b2 are parameters from Table 6, and
SC is the road surface coefficient from Table
7.
Table 6. The Coefficient of Rolling Resistance Tire
Parameters.
Tire Type b1 b2
Bias-ply
Standard Radial
Profile Radial
Wide Base Single Radial
0.636
0.424
0.350
0.303
0.00530
0.00495
0.00495
0.00495
Table 7. Road Surface Coefficient.
Road Surface Type
SC
Wet Black Top 0.8
Smooth Concrete 1.0
Worn Concrete, Brick, or Cold Black Top 1.2
Hot Black Top 1.5
Hard Packed Soil 1.5-2.0
Packed Gravel 12.0
Loose Gravel 7.5
Sand 12.0
Tire temperature also has a slight impact on rolling
resistance. Tire temperatures increase to their final
equilibrium operating temperature over a distance of
approximately 40 miles of travel. During the initial
40 miles of travel, rolling resistance is higher than
predicted by the above equation. The coefficient of
rolling resistance starts at about 120% of the value
achieved at equilibrium, and the drop-off is fairly
linear. Given the small impact that this factor will
have on overall engine load, the Phase I model will
assume that all vehicles have traveled 40 miles before
reaching the network. The Phase II model may
contain an enhancement if the phase I sensitivity
analysis indicates that tire temperature has a signifi-
cant effect on operating load.
Gravitational Weight Force (Fw)
The gravitational force components account for the
effect of gravity on vehicle weight when the vehicle
is operating on a grade. It is the component of the
vehicle weight parallel to the road. The grade angle
is positive on uphill grades (generating a positive
resistance) and negative on downgrades (creating a
negative resistance or a positive tractive force).
Fw = mx gx sin(<9)
(14)
Gravitational force of vehicle weight (/v) is ex-
pressed in units of pounds (force) and can be
re-written as a function of vehicle weight (W) and
road grade
Fw =
(15)
The combined vehicle weight includes the towing
vehicle plus the trailer load. The weight is not dis-
persed evenly throughout the combined vehicle, but
the units are assumed to always be operating on the
same grade.
To implement the gravitational force calculation in
the model, it is necessary to have information on road
grade. The relationship between Fw and 9 is fairly
straightforward. The sin(#) relationship, means that
an increase tractive force demand on a positive grade
exactly cancels a tractive force demand decrease on
the same negative grade. That is, if 50% of the
activity on a roadway link is on a 2% upgrade and
50% is on a 2% downgrade, the effect is the same as
if 100% of the activity had occurred on a flat road-
way. The effect is also nominally linear at grades
38
-------
Emissions Modeling Framework
experienced on major arterials and freeways in urban
areas. That is, the net effect of averaging grade on a
roadway link is not significant when the grades in
question are less than 10%. For a 60,000 pound truck
operating 50% of the time on a 10% grade and 50%
of the time on flat road, the net difference between
calculating each half separately and performing the
entire calculation at 5% grade is less than a difference
of 20 Ibf. This is less than a 0.4% difference in the Fw
calculation and is completely insignificant in terms of
the net change in tractive force when other factors
such as aerodynamic drag and roadway friction are
included in the calculations. Therefore, grade averag-
ing will be undertaken in implementing the gravita-
tional force algorithm on a link-by-link basis.
Although the effect of grade on power demand is
effectively linear and the average grade for that
roadway segment can be used in calculations that
employ a speed/acceleration matrix, this does not
mean that grade variability has the same linear effect
on on-road operations. For roadway segments that
contain variable grade, the on-road speed/acceleration
operations on the uphill segments can differ signifi-
cantly from the downhill segments. Similarly, opera-
tions on the extreme uphill segments will be different
than those on the level segments. In the example
provided above, a 5% grade can be used to reflect the
gravitational effect when half of the activity occurs at
10% grade and half is on level road. This is not the
case with the speed acceleration profiles. Grant
(1998) observed that the speed acceleration profiles
of trucks are significantly different on uphill grades
than on level terrain. In addition, Grant (1998)
observed that the presence of trucks on the uphill
freeway grades affects not only the truck speed/
acceleration conditions, but the speed/acceleration
conditions of the LDVs operating on the system as
well. Thus, the speed/acceleration matrices employed
in link-by-link calculations must also incorporate the
effect of grade.
Aerodynamic Drag Force (FD)
As a vehicle moves forward through the atmosphere,
drag forces are created at the interface of the front of
the vehicle and by the vacuum generated at the tail of
the vehicle. In fact, the flow of the air around the
vehicle creates a very complex set of force vectors
providing both resistance to forward motion as well
as vehicle lift. The sum of the drag forces is typically
expressed as aerodynamic drag (FD) in pounds
(force). Aerodynamic drag is a function of air den-
sity, drag coefficient (Q), vehicle frontal area, and
effective vehicle velocity.
(16)
where p is air density in pounds (mass) per cubic
foot,
Afis the vehicle frontal area in square feet,
and
Ve is the effective vehicle velocity in feet per
second.
At standard conditions (59 °F and 29.92 in of mer-
cury) air density is 0.076 lb/ft3. The air density, is a
function of atmosphere pressure and temperature, and
can be estimated in English units as
p= 0.07552
P.
_
519
29.92 A 460+ T.
(17)
where Pr is the atmospheric pressure in inches of
mercury, and
Tr is the atmospheric temperature in degrees
Fahrenheit.
In the load model, each HDV configuration will be
assigned a drag coefficient ranging from 0.6 to 0.99,
depending on the design and any aerodynamic
information that can be located in the literature.
Typical aerodynamic drag coefficients provided by
Ford Motor Company are shown in Figure 16.
Effective Vehicle Velocity (Ve) is the sum of vehicle
speed and wind speed, where wind speed is positive
if it blows toward the vehicle (headwind) and nega-
39
-------
Heavy-Duty Diesel Vehicle Model
Paractiute — >- ^Ej }
Fiat plate _ - H
Flat top _ . *,, *
tractor '4§"*JtPQ ~ no
High roof _ ^'^
sleeper W~~^9«i «w
Cone (60°; ~~*" yq
Hemisphere - (^
Thunderbird e?^"""3^}
(I«»4'-ai,:: "" O-
Cone (30°) ""j^J
Sphere j ]
Airfoil , • (,"~ " 1 7-~
1.35
1.17
0,99*
0 60"
0,51
0.41
0 35
0,34
0.10
0 05
Drag Coefficient of Various Shapes
'Source; Font Motor Co. 'National Research
Zot/ncil of Canada, **NASA)
Figure 16: Typical Aero-
dynamic Drag Coefficients.
tive if it blows in the direction of travel (tailwind).
Atmospheric winds vary in intensity, with typical
mean values of 10-20 mph. Given the nature of the
nonlinear relationship between effective velocity and
aerodynamic drag (a squared effect), ignoring on-
road winds (i.e., using a single average value of zero
for wind speed) will introduce a negative bias into the
predicted effect on engine load. Take, for example, a
single truck (i.e., constant drag coefficient and frontal
area) operating at a constant temperature (i.e., con-
stant air density). If this vehicle drives at 60 mph on
the freeway 50% of the time with a 15 mph headwind
and 50% of the time with a 15 mph tailwind, the net
drag force will be about 6% higher than if all travel
were undertaken with wind speeds equal to zero.
Introducing hourly wind fields into the model is
possible over the long term. However, link based
calculations would increase geometrically to reflect
the interactions between each vehicle configuration
and wind speed. In the Phase I model, a regional
wind rose will be used with the standard drag force
calculation equation to develop a correction factor
that will increase each drag force calculation by a
percentage to reflect the effect of variable wind
speeds and directions. Side wind velocities will be
ignored in the modal model, even though they do
contribute to vehicle yaw, which must be compen-
sated for through increased road load.
Drag forces do increase when a vehicle encounters
resistance air flow from a side angle. The flow of air
around the vehicle becomes uneven, with more drag
created on one side of the vehicle than the other.
Crosswinds increase the effective drag coefficient in
the relationship expressed above as a function of
wind angle. For example, a 10 degree angle of cross-
wind can increase the drag coefficient of a pickup
truck by 0.10 (Gillespie, 1992). In the Phase I model,
crosswinds will not be included.
Aerodynamic Lift Adjustment
The same flow of air across the vehicle that creates
an aerodynamic drag force also produces aerody-
namic lift. Lift force is proportional to air density,
vehicle speed squared, the coefficient of lift, and
frontal area. The lift generated by wind flow can
change the dynamic loading at each tire. The net
effect is to reduce the effective vehicle weight, which
will effectively reduce rolling resistance forces and
gravitational weight forces. Until reliable lift data can
be compiled for truck-trailer technologies, the lift
force will not be included in the modal model.
Curve Resistance Force (Fc)
Curve resistance is the additional frictional force on
the roadway associated with turning the vehicle and
resulting from the angled alignment of the front
wheels. The dynamic power loss can be estimated as:
0.002l5xV2xW
R.
(18)
40
-------
Emissions Modeling Framework
where Rm is the radius of curvature in feet.
This resistance force may not be significant on high
speed highways because the radius of curvature is
very large. Sensitivity analysis will be performed on
Atlanta freeways to determine if freeway estimation
can be ignored. However, curve resistance should be
relatively high at intersections when vehicles make
90 degree tight radius turns. For instance, the curve
resistance would be a large component of transit bus
tractive force, because the buses make numerous 90
degree intersection turns during daily service. Be-
cause bus routes are known commodities in the
GIS-based modeling system, transit vehicle modeling
will incorporate this function unless sensitivity
analysis determines that the effect is minimal and can
be ignored.
Payload Inertia! Resistance (FP)
When a vehicle or trailer is at rest, the load is distrib-
uted across the wheels such that the sum of the
moments about the center of mass cancel. That is,
when weight is placed toward the front of a trailer,
more of the weight is carried by the front axle and the
front tires. If the moments did not cancel, the trailer
would rotate about its center of mass.
When a vehicle and trailer accelerate, a new moment
is created, causing the vehicle and trailer to shift load
forces toward the rear axle. The front of the vehicle
and the trailer rotate upwards until the load on the
tires is sufficiently redistributed to stop the rotation.
Similarly, when a vehicle decelerates, the vehicle and
trailer rotate slightly forward and the load is shifted
toward the front axles and tires. A portion of the
motive force being delivered to the wheels is lost in
compensating for the rotational moment of inertia of
the vehicle and trailer load. Although the change in
rotational speed is small, changing from zero rota-
tions per second to a very slow rotational speed
caused by the temporary weight shift, the payload
involved may be quite large. The research team is
currently working through the calculation methodol-
ogy that can be implemented by vehicle and trailer
configuration type to incorporate the inertial loss
from the overall vehicle and trailer weights, as well
as from the spatial distribution of the payload within
the trailer. The Payload Inertial Loss (Fp) factor will
be included in the Phase II model for vehicle classifi-
cations and configurations for which the force is
significant.
Available Tractive Force (FT)
For "theoretically correct calculations," the dynamic
weight of the vehicle should include the effects of
acceleration, towing forces, curve forces, and even
the lift component of air resistance (Gillespie, 1992).
However, Gillespie (1992) also notes that introducing
some of the dynamic weight components in vehicle
performance estimation can complicate the calcula-
tions "without offering a significant improvement in
accuracy." In Phase II model development, sensitivity
analyses will help ascertain which dynamic compo-
nents should be added.
Any remaining tractive force after removing rolling
resistance force, gravitational weight force, and
aerodynamic drag force is available to accelerate the
vehicle. According to Newton's 2nd law, net remain-
ing force is equal to vehicle mass (ni) times accelera-
tion (a).
m x a = FT - FR - Fw - FD
W_
g
(19)
— xa=FT-FR-Fw-FD
The tractive force equation can be rewritten into the
power equation
p = '•
550
Fx (mx a+ FR + Fw + FD)
550
V x | — x a + FR + Fw + FD
(20)
W_
g
550
41
-------
Heavy-Duty Diesel Vehicle Model
For any vehicle and drive train technology groups
and for any given speed and acceleration combination
(from a speed/acceleration matrix for the technology
group), the tractive horsepower load can be esti-
mated. The first step is to calculate each of the
resistance coefficients using the relationships defined
earlier in this section. Using the average speed and
acceleration value in a matrix cell, the tractive horse-
power is calculated. Then, engine horsepower is
derived by solving for other horsepower losses
- p + p
rA T rDT
(21)
where PE is brake engine horsepower,
PDT is horsepower loss in the drive train, and
Pa is horsepower loss from accessory opera-
tions.
Hence, for every vehicle and drive train technology
group, and every speed/acceleration matrix cell, the
brake engine horsepower demand can be calculated.
42
-------
Emissions Modeling Framework
Emission Rate Functions
In the Phase I model, basic work-related emission
rates (grams per brake-horsepower-hour) will be
established for each technology group in the model.
As discussed earlier, technology groups are devel-
oped in an effort to group vehicle class, drive train,
and configuration combinations with similar labora-
tory performance and on-road load performance.
Appendix D provides the laboratories and contact
names from whom data have been requested. Appen-
dix E outlines some of the testing programs and data
that are available for development of the Phase I and
Phase II emission rate models.
Phase I modeling employs base emission rates (grams
per brake-horsepower-hour) for each technology
group derived from in-use engine testing certification
compliance data. Actual certification values will be
used for groups for which in-use testing data are
available to derive statistically significant emission
rates. The U.S. EPA and California Air Resources
Board (CARB) in-use observational databases will
provide these values.
The Phase II model will employ load-related emis-
sion rates. Regression analysis will provide the
emission rate relationship between grams per second
emission rate and axle-horsepower, as revealed
through preliminary analysis of chassis dynamometer
data (Ramamurthy and Clark, 1999; see Figure 17).
The data required for analysis must come from
chassis dynamometer and on-road test programs in
which second-by-second grams per second emission
rate data have been collected concurrently with
axle-horsepower loads. A linear or generalized
relationship is established between grams per second
06
04
0.3
0.2
01
y = -3(10"5 )x 2 + 0.0045* + 0 0397
R2 =0.6313
Vehicle B
y = -2(10"5)x2 +
-------
Heavy-Duty Diesel Vehicle Model
emission rate and tractive horsepower (axle horse-
power). Sufficient testing data are required to estab-
lish statistically significant samples for each technol-
ogy group.
Emitter Category
As discussed in the fleet characterization section,
on-road vehicles are classified as either high-emitters
or normal-emitters. The vast majority of HDVs are
normal emitters, but a small percentage of vehicles
have been tampered with or are poorly maintained.
For all on-road operations, on-road vehicle activity
by vehicle class, configuration, and drive train
technology will be divided into two emitter fractions.
Higher grams per brake-horsepower emission rates
will be applied to the high-emitter fraction based
upon review of the literature. Hence, for each tech-
nology group on the roadway, a high-emitter fraction
will be tracked. The applicable emission rates for
each technology group will be determined through
regression tree analysis of high emitting data (Wolf,
et al., 1998). For all on-road operations, on-road
vehicle activity by vehicle class, configuration, and
drive train technology will be divided into two
emitter fractions. Higher grams per brake-horse-
power-hour emission rates will be applied to the
high-emitter fraction, based on review of the litera-
ture. Hence, for each technology group on the road-
way, a high-emitter fraction will be tracked. In Phase
II model development, the goal will be to include a
hazard-type model to predict the fraction of high-
emitters on the road as a function of accrued vehicle
mileage, vehicle age, applicable inspection programs,
and so forth.
Commanded Fuel-Lean Operation
A significant percentage of on-road heavy-duty
trucks are subject to a recent EPA enforcement
action. The emissions from these vehicles were found
to be significantly lower in the laboratory under
standard certification test procedures compared to the
emissions noted under alternative testing conditions
that better reflect on-road operations. These engines
tended to jump into a lean-on-cruise (enleanment)
operating condition under extended cruise operations
on freeways, which provides significant fuel econ-
omy benefits for truck owners but significantly
increases emission rates for oxides of nitrogen.
Although many of the vehicles have been retrofitted
and reprogrammed to minimize the problem, recent
studies indicate that non-compliance may still be a
significant issue.
Information provided by the University of California
Riverside CE-CERT Mobile Heavy-Duty Vehicle
Laboratory (CE-CERT, 2004) illustrates the difficulty
in implementing a purely load-based model. The
"stylized" plot shown in Figure 18 illustrates NOX
emission rates vs horsepower.
Enleanment or "Cruise Points" -
Horsepower Remains Steady Over
Extended Period - Non-Linear Relationship*"
I
Non-Enleamnenl Events - Linear Relationship
Axel-Horsepower
Enleanment Events = High Air To Fuel Ratio Than Normal = Higher NOx Emissions
Typically, observed during high-speed "cruise" operations on freeways.
Non-Enleanment Events = Normal Air To Fuel Ratio = Reduced NOx Emissions
Typically, observed during lower-speed operations on roadways.
Figure 18. Stylized Plot of Axle-Horsepower vs
NOX Emissions (grams/second).
Although it is possible to examine the instantaneous
relationships between NOX emissions rates and axle
horsepower, the figure shows three "cruise" points
where the horsepower load remained steady for an
extended period. The activity represented by the
extended cruise conditions represents approximately
30% of the test data. When these data points are
removed from the data and placed in their own model
44
-------
Emissions Modeling Framework
regime, the remaining 70% of the data follow the
linear relationship between emission rate and load.
The first step toward modeling the effects of ex-
tended cruise is to establish the criteria under which
extended cruise emissions elevate. This can be
defined through analysis of laboratory testing results.
The definition would be established in terms of time
at speed and acceleration range. For example, labora-
tory testing may indicate for a vehicle class and drive
train technology that extended cruise begins once a
vehicle spends more than 45 seconds a single speed
(±5 mph) and acceleration rate (±1 mph/s) and ends
when the vehicle activity falls outside this window.
To incorporate the effect of extended cruise on
emissions, a third dimension will be added to the
speed/acceleration matrices for the vehicle class,
configuration, and technology groups affected by
extended cruise. This dimension will carry the per-
centage of activity in each speed/acceleration matrix
cell that occurs under extended cruise. Once in place,
this percentage of activity in each call can be as-
signed elevated emission rates appropriate for the
noted load.
Correction Factors and Environmental
Factors
The current modeling regime includes adjustments to
basic emission rates to account for the effects of
accrued vehicle mileage (deterioration), temperature,
humidity, and altitude. The basic modeling ap-
proaches employed in the current emission rate
models are outlined in Appendix F. In the MOBILE6
modeling regime, all correction factors are assumed
to have independent effects on basic emission rates.
The same will be true in this modeling regime,
sufficient data are collected with adequate controls
over multiple variables such that interaction effects
can be determined. Each correction factor is dis-
cussed in turn.
Deterioration Rates
As LDVs age and accrue vehicle miles of travel,
emission rates (gram per hour or gram per mile) tend
to increase. Evidence of the deterioration of vehicle
combustion and control systems are evident in the
LDV fleet through in-use laboratory testing pro-
grams, inspection and maintenance testing programs,
and remote sensing programs. The current MOBILE6
Model includes a deterioration rate effect on emis-
sions. However, based upon review of the emission
testing data used in the development of the EMFAC-
2000 motor vehicle emission factor model, the
increase in emissions rates over time were not ade-
quately demonstrated within the heavy-duty diesel
truck test fleet used to develop the EMFAC2000
emission rates.
In developing EMFAC2000, CARB compiled chassis
dynamometer test results for medium-heavy-duty
(14,001 to 33,000 pounds GVWR) diesel vehicles
(MHDDVs) and heavy-heavy-duty (more than 33,001
GVWR) diesel vehicles (HHDDVs).Three data sets
were available for HHDDVs (14 vehicles tested for
New York by West Virginia University, 5 vehicles
tested at high-altitude by the Colorado School of
Mines, and 5 additional vehicles tested by West
Virginia University). MHDDV tests included 21
vehicles tested for New York by West Virginia
University and 6 vehicles tested at high-altitude by
the Colorado School of Mines. All vehicles ranged
from 1981 to 1998 model year. Each vehicle was
tested from two to six times (replicate testing), and
test results were averaged for each vehicle.
In assessing the West Virginia testing data employed
in EMFAC2000 development (leaving out the high-
altitude tests due to problems noted in the altitude
correction factor discussion that follows), dummy
variables were created for engine certification model
year groups (i.e., when different certification stan-
dards for various pollutants applied to each group of
engines). The dummy variables serve as surrogates
for changes in emissions control systems and engine
computer algorithms that may have helped reduce
emissions. Certification groupings were pre-1984,
1984-1987, 1988-1990, 1991-1993, 1994-1997, and
1998-2002. Interact!on variables were created for the
certification groups to test the interactions of these
groups with odometer reading. From a theoretical
45
-------
Heavy-Duty Diesel Vehicle Model
perspective, use of certification group coupled with
deterioration interactions is preferable to emissions
derived as a function of model year.
Regression analysis results indicated that modeling
PM emissions is significantly improved when mod-
eled as a function of certification groups rather than
model year. Accrued vehicle mileage was not a
significant explanatory variable, probably because the
accrued mileage effect is already partly explained by
the certification group (i.e., is correlated with vehicle
model year).
Based on the limited data examined, deterioration
cannot be differentiated from vehicle age as reflected
in the certification group. Heavy-duty diesel engines
are sleeved for rebuild, so it is possible that engine
overhauls minimize any deterioration effect within a
certification group. Larger samples would likely
support the development of accurate deterioration
rates.
A significantly expanded data set, tested under a
wider variety of conditions and in both altitude
locations, would help determine the factors likely to
dominate the emissions effect. In updating the heavy-
duty emission rates for upgrading EMFAC2000 to
EMFAC2002, CARS employed 75 engine tests
(CARS, 2002). These data will be procured and
analyzed in the same manner. The Energy and Envi-
ronmental Analysis, Inc. (EEA, 2000) report that
provides the basis for the deterioration rates will be
critically reviewed as well. In the Phase I model,
deterioration rates will not be included. Deterioration
rates will be incorporated into the Phase II model if
statistically significant effects can be determined
from data analysis.
Temperature
In the MOBILE6 model, temperature correction
factors adjust exhaust emissions to temperatures that
fall outside of the standard laboratory conditions
(within a window surrounding 75 °F). Tests are
performed on vehicles or engines at a variety of
temperatures, and the ratios of observed test results
and baseline test results establish the relationship
between emission rates and temperature. Temperature
correction factors are not applied to diesel vehicles,
so from the perspective of model implementation this
is not a problem.
High ambient temperature can also increase engine
load due to increased air conditioner use. If data are
available to provide relationships between tempera-
ture and humidity and air conditioning usage, these
relationships will be integrated into the accessory
load module in the Phase II model.
Humidity
Data used to develop humidity correction factors
have not yet been compiled and assessed. Previous
assessments performed by the research team indicate
that the correction factors are likely based on small
samples (and may notbe statistically significant). The
test results will be compiled and reassessed. If the
correction factors are defensible, they will be incor-
porated into the model as a linear adjustment factor.
Altitude
MOBILE6 employs a high-altitude HDV emission
rate correction factor of 1.47, based on EPA's report
prepared during the development of MOBILE6
(CARB, 2000). Although this correction could be
built directly into the model and applied to all tech-
nology groups and operating conditions, the research-
ers believe that this should not be done until such a
relationship is clearly established through statistical
analysis. Based on review of the emission testing data
used in the development of EMFAC2000, the in-
crease in emissions rates over time were not ade-
quately demonstrated within the test fleet.
In assessing the West Virginia and Colorado (high-
altitude) test data employed in EMFAC2000 develop-
ment, dummy variables were created for engine
certification model year groups (pre-1984, 1984-
1987, 1988-1990, 1991-1993, 1994-1997, and 1998-
2002), and interaction variables were created for the
certification groups to test the interactions of these
groups with testing at high altitude. Altitude was
46
-------
Emissions Modeling Framework
entered into the equation as a dummy variable, and
the Colorado data appeared to show altitude effects
with respect to PM emissions. This impact appeared
to influence the intercept term of the equation.
However, when modeled as an interaction variable
across the certification groups, the analyses indicated
that altitude interactions were only present for some
(older) certification groups. The findings indicated
that altitude corrections should probably be modeled
separately across certification groups.
Larger data sets will be required to make this deter-
mination. The data set employed in the development
of EMF AC2002 will be similarly reviewed. A remote
sensing study also appears to indicate that there is a
significant relationship between altitude and HDV
performance (Bishop, 2001). Bishop reported the
remote sensing measurements of emissions from
5772 heavy-duty diesel trucks between 1997 and
1999 at five locations in the United Sates and Europe.
The results show a statistically significant increase in
carbon monoxide, hydrocarbons, and nitric oxide
with altitude. The report also indicates an increase in
fuel consumption as well (Bishop, 2001). This study
will be reviewed by the research team.
The Phase I model will not incorporate an altitude
correction factor (which, in any case, is not needed
for Atlanta). If the results from the review of the
EMFAC2002 database and the Bishop (2001) study
indicate that high-altitude correction factors are
statistically defensible, they will be integrated in the
Phase II model as a linear adjustment factor by
pollutant.
47
-------
Heavy-Duty Diesel Vehicle Model
48
-------
Emissions Modeling Framework
Inventory Assembly and Model Output
The inventory assembly process is essentially a set of
link-by-link emissions calculations using the link
attributes, subfleet technology group attributes, and
the load-based equations. The calculations are pro-
cessed as follows:
Emission Matrix Calculations
• Speed/Acceleration Matrix - For each vehicle
class, configuration, and drive train technology
group, an applicable speed/acceleration matrix is
selected.
• Total Link Vehicle Hours - Given the technology
group traffic volumes (vehicles per hour travers-
ing the link), the road length, and average travel
speed for the road, total technology group
vehicle-hours of travel are computed for the link.
• Vehicle-Hour Matrix - The speed and accelera-
tion frequency distributions provide the fraction
of on-road hours of travel that are undertaken in
each speed/acceleration bin. Multiplying the
matrix by total vehicle-hours of travel provides
the vehicle-hours of travel for each speed/accel-
eration bin.
• Road Load Matrix - The road load power demand
(brake-horsepower) associated with each speed/
acceleration matrix cell can be calculated using
(1) the average speed and acceleration rate in
each matrix cell and (2) the various vehicle and
roadway parameters from each vehicle class and
configuration technology group.
• Inertial Loss Matrix - The inertial loss matrix
associated with specific drive train technology
provides the inertial power loss for each matrix
cell.
• Accessory Load Matrix - The accessory load
matrix provides the accessory power loss for each
matrix cell.
• Power Demand Matrix - The values in the road
load matrix are added to the technology group
inertial loss matrix and accessory load matrix to
estimate total engine power demand (brake-
horsepower) for the specific operating conditions
in that cell.
Phase I Model Emission Calculations:
• Work Matrix - The power demand matrix (brake-
horsepower in each cell) is multiplied by the
vehicle-hour matrix (hours in each cell) to obtain
a matrix of brake-horsepower-hour engine work
by speed/acceleration matrix cell.
• Average Work-Related Emission Rate - For each
technology group, the average work-related
emission rate (grams per brake-horsepower-hour)
from dynamometer testing is quantified.
• Emission Calculation - The work matrix is multi-
plied by the average work-related emission rates
for the technology group (grams per brake-horse-
power-hour).
• Cell Addition - Total grams per cell are added to
develop the hourly total emissions (grams per
hour).
Phase II Model Emission Calculations:
• Emission Rate Matrix - Emission rates (grams
per second) are modeled as a function of brake-
horsepower load. Using the values in the power
demand matrix (brake-horsepower) for each cell,
a matrix of emission rates (grams per second) is
created for each technology group and converted
to grams per hour, where the emission rate in
each cell applies to the activity in that cell.
• The emission rate matrix values are multiplied by
their counterparts in the vehicle-hour matrix to
estimate grams of emissions for each matrix cell.
49
-------
Heavy-Duty Diesel Vehicle Model
• Cell Addition - Total grams per cell are added to
develop the hourly total emissions (grams per
hour).
Integration of Link-Based Emissions
Working in the GIS system and using a graphic user
interface similar to the one developed for MEASURE
(Figure 19), the user can either examine link-by-link
emissions (grams per link per hour) or aggregate the
emissions predictions to any grid cell size. Because
the hourly predicted emissions become link attributes
in the GIS system, the user can create useful graphics
to illustrate the source and intensity of heavy-duty
diesel emissions. Figure 20 illustrates the graphic out-
put capabilities of the MEASURE model, and similar
capabilities will be integrated into the Heavy-Duty
Modal Model (Bachman, 1997). Because the emis-
sions are retained at the link level and also aggre-
gated on a grid cell basis, the emissions predictions
can be used in line source microscale impact assess-
ment as well as in regional ozone or particulate
matter dispersion modeling.
ATLANTA AREA MOBILE EMISSIONS
SCCOMO CCNCMTI0N MOQ1C EMISSION PnC'jCCT
HT MOFMILF
Figure 19. Graphic User Interface Developed for MEASURE.
50
-------
Emissions Modeling Framework
Figure 20. Example of Grid 8-9 AM Cell Emissions from MEASURE.
51
-------
Heavy-Duty Diesel Vehicle Model
52
-------
Emissions Modeling Framework
Case Study of Two MARTA Transit Bus Routes
In late summer 2004, the research team equipped two
MARTA transit buses with the same Georgia Tech
trip data collectors used in the Commute Atlanta
instrumented vehicle study. Second-by-second data
were collected from these two vehicles over a period
of months while the buses served on a wide variety of
metro routes. The MARTA bus SAFD is truncated at
55 mph (rather than extending to 100 mph) because
speeds higher than 50 mph do not take place on the
city routes. Acceleration bins in the example matrix
range from -15 to +15 ft/s2, in 1 ft/s2 bins. Accelera-
tions that are greater than 15 ft/s2 or less than -15 ft/s2
are placed in the last bin. Cells on speed-acceleration
matrices were filled with acceleration frequency
fractions for corresponding speed and acceleration
bins. For each roadway type and for each time of day,
a unique speed-acceleration matrix can be created.
The months of second-by-second on-road bus operat-
ing data collected during the late summer of 2004
was processed to develop matrices for use in load
modeling. Figure 21 is an example of some of the
MARTA routes sampled in Atlanta over a three week
period. In order to secure data samples from a variety
of buses, the Georgia Tech research team also in-
stalled a trip data collector on a MARTA transit bus
different from the ones used in the late summer 2004
program and collected second-by-second speed data
for three weeks, from June 28 to July 17, 2004, as
part of a six month data collection program. Data are
collected during all times of vehicle operation, as the
bus travels MARTA service routes and during dead-
head operation. Transit bus speed and location data
collected with the GPS receiver and stored in the data
storage are remotely transmitted to a server computer
managed by Georgia Tech. From the second-by-
second speed data, researchers calculate correspond-
ing second-by-second acceleration. Then, speed and
acceleration data are grouped by roadway type and by
time of day and used to create road and time-of-day
specific speed/acceleration matrices.
After the map matching process, transit bus trips for
three weeks were identified on ArcGIS, a desktop
mapping product by Environmental Systems Re-
search Institute (ESRI). From the transit bus loca-
tions, four types of activities were observed: regular
bus service, approach for service, return after service,
and idling at garages. Among the four types of
services, only the regular bus service activity was
considered for the speed and acceleration analyses.
During the three-week study period, the bus served
ten regular bus routes for more than fifteen vehi-
cle-days total. The data points on the bus service
routes provided 249,022 seconds of activity. Among
the data points, 64% of data were on arterial roads,
35% on local roads, and 1% on freeways. Relatively
few data points were observed on freeways, so they
were excluded from analyses. Researchers removed
data points with speed of zero mph (representing
idling at intersections, bus stops, and terminal) before
beginning speed and acceleration analyses on arterial
and local roads. In total, 244,203 data points (157,471
on arterial roads and 86,732 on local roads) were
analyzed for transit bus speed and acceleration
characteristics. The data clouds are provided in
Figures 22 and 23, while Figures 24 and 25 provide
the data in traditional Watson plots format. Finally,
the binned results can be employed as a calculation
matrix described previously (Figure 26).
53
-------
Heavy-Duty Diesel Vehicle Model
Figure 21. Example MARTA Transit Bus Routes on which Speed and Location Data
were Collected.
54
-------
Emissions Modeling Framework
10.00-
fc. 0.00-
o
01
•S -5.00'
•10.00-
Motning
I I
0.00 10.00 2000 30.00 40.00 50.00 6000 7000
Speed (mph)
10.00-
500-
c
•c o.oo-
I
"5
-5.00-
-10.00-
MWday
I I I
0.00 10.00 20.00 3000 4000 50.00 60.00 70.00
Speed(mph)
10.00-
5.00-
f
fi
o
HI
o
0.00-
-5.00-
-10.00-
Afteinoun
10.00-
5.00-
i
£-
-------
Heavy-Duty Diesel Vehicle Model
10.00-
5.00-
•& o.oo-
u
i
-5.00-
-10.00-
Muininq
I I I I I I
0.00 10.00 20.00 30.00 4000 50,00 GO .00 70.00
Speed(mph)
10.00-
5.00-
0.00-
u
<
-5.00-
-10.00-
Mdtlay
0.00 10.00 20.00 30.00 40.00 50.00 GOOD 70.00
Speed (mph)
10.00-
5.00-
o
o
3
0.00-
-5.00-
-10.00-
Afternoon
1000-
5.00-
0.00-
-5 00 -
-1000-
I I
0.00 10.00 20.00 30.00 40.00 50.00 SO.OO 70.00
0.00 10.00 2000 30.00 40.00 50.00 60.00 70.00
Speed (mph)
Speed (mph)
Figure 23. Speed-Acceleration Scatter Plots for Local Roads by Time of Day.
56
-------
Emissions Modeling Framework
i q p •
• o o
I • D
p p p q q •
o o o o o
D • G C • :
s
a
o 2
•O <5
If
to 3.
co
I
S
._
OOl«l^tfl«l^rrt(MT-
T-£>oo5pqqoo
''
ra
111
m
I
o o
< £
"•••DBDBDDBa
2gg|gSSgSq
0000000000
• • D • D•G D••
Figure 24. Speed-Acceleration Profiles for Arterial Roads by Time of Day.
57
-------
Heavy-Duty Diesel Vehicle Model
li
pppp
c* + o
pop
12
o °. .
(Q 00
£ • •
ppppppppp
'oooooooo
• r • • • n
OpOOOOOOOO
0600060600
BBDBSBDOBn
SS&gSSSSs:
• •LJBDIL:H
Figure 25. Speed-Acceleration Profiles for Local Roads by Time of Day.
58
-------
Emissions Modeling Framework
Speed (mph) Bins
10 25 30
-
0.000
0.001
0.003
0.005
0.005
0.010
0.090
0.064
0.096
0.011
0.003
0.000
0.001
0.001
0.002
0.002
0.003
0.002
0.003
0.002
0.002
0.004
0.011
0.003
0.000
.
-
0.000
0.000
0.00!
0.003
0.004
0.005
0.006
0.005
0.004
0.005
0.000
0.006
0.006
0.013
0.016
0.001
„
0.000
0.000
0.000
0.000
0.002
0.003
0.005
0.007
0.007
0.007
0.007
0.000
0.010
0.011
0.014
0.013
0.004
0.000
.
-
0.000
0.000
0.000
0.001
0.003
0.004
0.006
0.007
0.009
0.011
0.001
0.013
0.019
0.020
0.008
0.002
0.000
.
-
0.000
0.000
0.001
0.001
0.003
0.005
0.007
0.010
0.020
0.001
0.032
0.020
0.015
0.004
0.001
0.000
.
-
0.000
0.000
0.000
0.001
0.001
0.003
0.005
0.010
0.033
0.002
0.047
0.022
0.008
0.002
0.000
.
-
0.000
0.000
0.001
0.001
0.002
0.006
0.025
0.001
0.037
0.013
0.002
0.001
0.000
.
-
0.000
0.000
0.001
0.004
0.015
0.001
0.018
0.006
0.001
0.000
0.000
.
-
0.000
0.000
0.001
0.004
0.000
0.006
0.002
0.000
.
-
0.000
0.000
0.000
0.000
0.000
0.000
0.000
.
-
0.000
0.000
.
Figure 26. Example of a Speed-Acceleration Matrix (Arterial Road, Morning Peak Period).
59
-------
Heavy-Duty Diesel Vehicle Model
Required horsepower for each speed bin can be
weighted by acceleration frequency fractions on
corresponding speed-acceleration bins from the
matrix, and weighted required horsepower are aggre-
gated as a unique required horsepower for the se-
lected roadway type and by time of day. Then, the
unique required horsepower is multiplied by each bus
model year emissions level in grams per brake-
horsepower-hour (in this case the 4.0 g/bhp-hr certifi-
cation rate was used), to calculate an emissions in
grams per hour for the selected bus service route. In
this example, inertial and accessory loads are ig-
nored.
where EM is the transit bus emissions in grams per
hour per vehicle,
EL is the transit bus emissions rate in grams
per brake-horsepower-hour,
P is the engine power demand in brake-
horsepower-hour,
AF is the acceleration/deceleration activity
frequency,
/' is the roadway type (arterial or local road),
j is the time of day (morning, midday, after-
noon, or night),
k is the engine model year,
/is the speed in a speed/acceleration matrix,
and
m is the acceleration in a speed/acceleration
matrix.
To demonstrate emissions differences between
different operating speeds, two cells (7.5 and 37.5
mph at +1 mph/s), which have same acceleration
frequency fractions (0.009), were selected from the
speed-acceleration matrix for morning arterial roads,
and load-based transit bus required horsepower was
calculated. Required acceleration forces at 7.5 and
37.5 mph are the same. However, the sum of the
other forces (rolling resistance, gravitational drag,
and aerodynamic drag) at 37.5 mph was 2.2 times
greater than at 7.5 mph. This is because aerodynamic
and rolling resistance drags increase as vehicle speed
increases. Total engine power demand estimated at
37.5 mph was 5.7 times greater than at 7.5 mph; that
implies that transit bus emissions at 37.5 mph will be
5.7 times greater than at 7.5 mph for +1 mph/s
acceleration at the condition of linear relationships
between emissions level and engine power demand.
Demonstration of the differences between emission
estimates using this modeling approach vs MOBILE6
will be provided in Phase II of the project.
The net difference in grams per hour for operation on
each of the routes is significant. The second-by-
second road loads for vehicles operating on two
different routes are primarily a function of speed and
acceleration and grade (Figure 27). Using the load-
based estimation tools, buses traveling on Route 13
are predicted to emit approximately 279 g/h and the
same buses traveling on Route 23 are predicted to
emit approximately 404 g/h. The difference in grams
per hour emission rates is approximately 45%. Given
that the average sped on Route 13 is 11 mph and the
average speed on route 23 is 15 mph, the net differ-
ence in grams per mile emission rates is approxi-
mately
22%.
The spreadsheet transit modal model provides the
input necessary to complete most of the algorithms in
the ArcGIS framework. Over the next few months,
the coding of the Phase I model will be completed in
the GIS system, so technology groups can be tracked
and tabulated within the model as they are developed.
60
-------
Emissions Modeling Framework
1100
1050
LU
700
Bus Route 13 Required HP vs Elevation
1200
T- T- CM CO
Time (second)
Bus Route 23 Required HP vs Elevation
1100
1050
Time (second)
Figure 27. Load Calculations for two Bus Routes with Different Grades and
Operating Profiles.
1200
-------
Heavy-Duty Diesel Vehicle Model
62
-------
Emissions Modeling Framework
References
Ahanotu, D. (1999). Heavy-Duty Vehicle Weight and
Horsepower Distributions: Measurement of Class-Specific
Temporal and Spatial, Thesis, Georgia Institute of Tech-
nology.
Bachman, W. (1997). A GIS-Based Modal Emissions
Model, Thesis, Georgia Institute of Technology.
Bachman, W., W. Sarasua, S. Hallmark, and R. Guensler
(2000). Modeling Regional Mobile Source Emissions in a
GIS Framework, Transportation Research Part C:
Emerging Technologies, 8:1-6, pp. 205- 229.
Bishop, G.A., J.A. Morris, D.H. Stedman, L.H. Cohen,
R.J. Countess, P. Maly, S. Scherer. (2001). The Effects of
Altitude on Heavy-Duty Diesel Truck On-Road Emis-
sions, Environmental Science and Technology, 35:8, pp
1574-1578.
CAEC (2004). Cleaire Advanced Emission Controls, LLC
http://www.cleaire.com/ (accessed August 2005).
CARB (2000). Public Meeting to Consider Approval of
Revisions of State's On-road Motor Vehicles Emissions
Inventory Technical Support Document, California Air
Resources Board.
CARB (2002). EMFAC(2002)
http://www.arb.ca.gov/msei/on-road/latest revisions.htm#hhddt idle
(accessed August 2005), California Air Resources Board.
CE-CERT (2004). Personal Communication. Center for
Environmental Research and Technology, University of
California, Riverside http://www.cert.ucr.edu/ (accessed
August 2005).
Clark, N. and M. Gautam (2004). Medium Heavy-Duty
Truck Test Cycle Evaluation, Publication No. CRC Proj ect
No. E-55-3. Coordinating Research Council.
Colorado IFHAER (2004). Colorado Institute for Fuels
and High Altitude Engine Research, Colorado School of
Mines http://www.mines.edu/research/cifer/ (accessed
August 2005).
EEA (2000). Documentation and Analysis ofHeavy-Duty
Diesel Vehicle Emission Test Data, Prepared for New
York State Department of Environmental Conservation
(NYSDEC), Energy and Environmental Analysis, Inc.
FHWA (1994). Bridge Formula Weights, Publication No.
FHWA-MC-94-007. U.S. Department of Transportation.
FHWA (2000). Highway Statistics, Series, (1993 to 1999),
U.S. Department of Transportation.
FHWA (2001). Traffic Monitoring Guide, Publication No.
FHWA-PL-01-021 U.S. Department of Transportation.
FHWA (2002). Truck Size and Weight, Route Designa-
tions: Weight Limitations (23CFR658.17). U.S. Depart-
ment of Transportation.
FHWA (2004). Highway Statistics Series (2003), U.S.
Department of Transportation.
Gautam, M. and N. Clark (2003). Heavy-Duty Vehicle
Chassis Dynamometer Testing for Emissions Inventory,
Air Quality Modeling, Source Apportionment and Air
Toxics Emissions Inventory, Publication CRC Project No.
E-55/E-59. Coordinating Research Council, Inc.
Gillespie, T. (1992). Fundamentals of Vehicle Dynamics,
ISBN 1-56091-199-9, Society of Automotive Engineers,
Inc., Warrendale, PA.
Grant, C. (1998). Modeling Speed/Acceleration Profiles
on Freeways, Thesis, Georgia Institute of Technology.
Guensler, R., D. Sperling, and P. P. Jovanis (1991).
Uncertainty in the Emission Inventory for Heavy-Duty
Diesel-Powered Trucks, Proceedings for the 84th Air and
Waste Management Association Annual Meeting, Pitts-
burgh, PA.
Guensler, R., M. Thornton, M. O. Rodgers, K. Dixon, J.
63
-------
Heavy-Duty Diesel Vehicle Model
Pearson, W. Bachman, J. Leonard, and J. Daniel (2001).
Evaluation of Ramp Metering Impacts on Air Quality: The
Atlanta 1-75 Case Study, Georgia Transportation Institute,
Georgia Institute of Technology.
Guensler, R., K. Dixon, V. Elango, and S. Yoon (2004).
MOBILE-Matrix: Georgia Statewide MTPT Application
for Rural Areas, Transportation Research Record: Journal
of Transportation Research Board, No. 1880, Transporta-
tion Research Board, National Research Council, Wash-
ington, DC, pp. 83-89.
Heywood, J.B. (1988). Internal Combustion Engine
Fundamentals, McGraw-Hill Publishing Co., New York,
NY.
Lindhjem, C., and T. Jackson (1999). Update of Heavy-
Duty Emission Levels (Model Years 1998-2004+) for Use
in MOBILE6, EPA-420/R-99/010, Office of Transporta-
tion and Air Quality, U.S. Environmental Protection
Agency, Washington, DC.
NREL (2004). National Renewable Energy Laboratory,
U.S. Department of Energy
http://www.nrel.gov/vehiclesandruels/data performance.html
(accessed August 2005).
Ramamurthy, R. andN. Clark (1999). Atmospheric Emis-
sions Inventory Data for Heavy-Duty Vehicles, Environ-
mental Science and Technology, 33:1, pp 55-62.
Rodgers, M., R. Guensler, J. Pearson, O. Kemenova, S.
Yoon, and P. Zhang (2004). Atlanta Heavy-Duty Vehicle
and Equipment Inventory and Emissions Study
(AHDVEIES): Data Collection Report, Submitted to
Georgia Regional Transportation Authority, Georgia
Institute of Technology.
SAE (2004). Information Relating to Duty Cycles and
Average Power Requirement of Truck and Bus Engine
Accessories, SAE J1343 (August 2000). Society of
Automotive Engineers, Warrendale, PA.
U.S. EPA (1998). Update of Fleet Characterization Data
for Use in MOBILE6-Final Report, EPA-420/P-98/016,
Office of Transportation and Air Quality, U.S. Environ-
mental Protection Agency, Washington, DC.
U.S. EPA (2002b). User's Guide to MOBILE6.1 and
MOBILE6.2: Mobile Source Emission Factor Model,
EPA420/R-02/028, Office of Transportation and Air
Quality, U.S. Environmental Protection Agency, Washing-
ton, DC.
U.S. EPA (2004). National Vehicle and Fuel Emissions
Laboratory, Office of Transportation and Air Quality, U.S.
Environmental Protection Agency, Washington, DC,
http: //www. epa.gov/nvfel/ (accessed August 2005).
Wolf, J., R. Guensler, and S. Washington (1998). High
Emitting Vehicle Characterization Using Regression Tree
Analysis, Transportation Research Record: Journal of
Transportation Research Board, No .1641, Transportation
Research Board, National Research Council, Washington,
DC, pp. 58-65.
Yoon, S., P. Zhang, J. Pearson, R. Guensler, and M.
Rodgers (2004). A Heavy-Duty Vehicle Visual Classifica-
tion Scheme: Heavy-Duty Vehicle Reclassification
Method for Mobile Source Emissions Inventory Develop-
ment. Proceedings of the 97th AWMA Annual Meeting.
Indianapolis, IN.
Yoon, S., M. Rodgers, J. Pearson, and R. Gansler (2004b).
Engine and Vehicle Characteristics of Heavy-Duty Diesel
Vehicles in the Development of Emissions Inventories:
Model Year, Engine Horsepower and Vehicle Weight,
Transportation Research Record: Journal of Transporta-
tion Research Board, No. 1880, Transportation Research
Board, National Research Council, Washington, DC, pp.
99-107.
Yoon, S., H. Li, J. Ogle, R. Guensler, and M. Rodgers.
(2005a). A Methodology for Developing Transit Bus
Speed-Acceleration Matrices to be Used in Load-Based
Mobile Source Emissions Models, Transportation Re-
search Record: Journal of Transportation Research
Board, Transportation Research Board, National Research
Council, Washington, DC, (in Press).
Yoon, S. (2005b) A New Heavy-Duty Vehicle Visual
Classification and Activity Estimation Method for Re-
gional Mobile Source Emissions Modeling. Thesis.
Georgia Institute of Technology.
VIUS (2002). Vehicle Inventory and Use Survey. U.S.
Census Bureau.
http: //www .census. gov/svsd/www/vius/2002 .html
(accessed August 2005).
64
-------
Emissions Modeling Framework
Bibliography
Bachman, W., J. Granell, R. Guensler, and J. Leonard
(1998). Research Needs In Determining Spatially Re-
solved Subfleet Characteristics. Transportation Research
Record: Journal of Transportation Research Board, No.
1625, Transportation Research Board, National Research
Council, Washington, DC, pp. 139- 146.
BOSCH (1986). Automotive Handbook, 2nd Edition,
VDI-Verlag, GmbH (ISBN 3-18-418006-9).
Brieman, L., J.H. Friedman, R.A. Olshen, andC.J. Stone
(1984). Classification and Regression Trees, Wardsworth
International Group, Belmont, CA.
CARB (1998). Methodology for Estimating Emissions
from On-road Motor Vehicles, Volume II: EMFAC7G,
California Air Resources Board.
Census Bureau (2004). 2002 Vehicle Inventory and Use
Survey, Georgia, U.S. Census Bureau.
Fischer, M., M. Han, F. Baron, R. Capelle, L. Chimini, R.
Czerniak, T. Dahlburg, A. Danaher, S. Kale, T. Palmerlee,
C. Sanft, R. Snyder, C. Walters (2001). Truck Trip
Generation Data, A Synthesis of Highway Practice,
NCHRP Synthesis 298. Transportation Research Board.
Washington, DC.
Fomunung, I., S. Washington, and R. Guensler (1999). A
Statistical Model for Estimating Oxides of Nitrogen
Emissions from Light-Duty Motor Vehicles, Transporta-
tion Research Part D: Transport and Environment, 4:5,
pp. 333-352.
Fomunung, I., S. Washington, R. Guensler, and W.
Bachman (2000). Validation of the "MEASURE" Auto-
mobile Emissions Model—A Statistical Analysis. Journal
of Transportation Statistics; 3:2, pp. 65-84.
Frey, C. and J. Zheng (2002). Probabilistic Analysis of
Driving Cycle-Based Highway Vehicle Emission Factors,
Environmental Science and Technology, 36:23, pp. 5184-
5191.
GA DNR (2004). State Implementation Plan for the
Atlanta 1-Hour Ozone Nonattainment Area, Georgia
Department Natural Resources, (http://www. gaepd.org/)
(accessed August 2005).
Garg, V. and R Dukkipati (1984). Dynamics of Railway
Vehicle Systems, ISBN 0-12275-950-8, Academic Press
Canada.
Guensler, R., W. Bachman, and S. Washington (1998).
An Overview of the MEASURE GIS-Based Modal
Emissions Model, Transportation Planning and Air
Quality III, Tom Wholley, Ed., American Society of Civil
Engineers. New York, NY. pp. 51- 70.
Guensler, R. and J. Ogle (2003). Atlanta's Vehicle Instru-
mentation and Activity Monitoring Programs: 2003 Status
Report, presented at the 13th CRC On- Road Vehicle
Emissions Workshop. San Diego, CA.
Hallmark, S. (1999). Analysis and Prediction of Individ-
ual Vehicle Activity for Microscopic Traffic Modeling,
Thesis, Georgia Institute of Technology.
Hallmark, S., R. Guensler, and I. Fomunung (2002).
Characterizing On-road Variables that Affect Vehicle
Modal Operation, Transportation Research Part D:
Transport and Environment, 7:2, pp. 81- 98.
Hay, W. (1982). Railroad Engineering, Second Edition,
John Wiley & Sons, Inc., Canada.
Jimenez-Palacios, J. (1999). Understanding and Quantify-
ing Motor Vehicle Emissions with Vehicle Specific Power
and TILDAS Remote Sensing, Thesis, Massachusetts
Institute of Technology.
65
-------
Heavy-Duty Diesel Vehicle Model
Miller, W., Cocker, D., Johnson, K., Norbeck, J.M., Park,
C.S., and Welch, B. (2002). Use of a Mobile On-Road
Laboratory to Measure HDD 'Real World' Emissions
from Standard and Non-Standard Operating Cycles,
presented at the 12th CRC On-Road Vehicle Emissions
Workshop, San Diego, CA.
Nam, E. (2003). Proof of Concept Investigation for the
Physical Emissions Rate Estimator (PERE) for MOVES,
EPA-420/R-03/005, Office of Transportation and Air
Quality, U.S. Environmental Protection Agency, Washing-
ton, DC.
Russell, A., Bailar III, J., Earth, M., Caretto, L., Howard,
C., Johnson, J., Kowalczyk, J., Lloyd, A., Morris, M.,
Pollack, A., and Sawyer, R (2000). Modeling Mobile-
Source Emissions, Publication No. 0-309-07088-0,
National Research Council. Washington, DC.
SAE (1992). Truck Systems Design Handbook, Publication
No. SAE PT-41, Society of Automotive Engineers, Inc.,
Salem, MA.
Sawyer, R, Harley, R, Cadle, S., Norbeck, J., Slott, R.,
and Bravo, H. (2000). Mobile sources critical review:
1998 NARSTO assessment, Atmospheric Environment,
34:12-14, pp. 2161-2181.
U.S. EPA (1995). Draft User's Guide to PARTS: A
Program for Calculating Particle Emissions from Motor
Vehicles, EPA-AA-AQAB-94-2, Office of Mobile
Sources, U.S. Environmental Protection Agency, Wash-
ington, DC.
U.S. EPA (2001). EPA Requirements for Quality Assur-
ance Project Plans, EPA QA/R-5, EPA-240/B-01/003,
Office of Environmental Information, U.S. Environmental
Protection Agency, Washington, DC.
U.S. EPA (2002a). Guidance for Quality Assurance
Project Plans for Modeling, EPA QA/G-5M, EPA-
240/R-02/007, Office of Environmental Information, U.S.
Environmental Protection Agency, Washington, DC.
U.S. EPA (2003). MOBILE6.1 Particulate Emission
Factor Model Technical Description Final Report, EPA-
420/R-03/001, Office of Transportation and Air Quality,
U.S. Environmental Protection Agency, Washington, DC.
66
-------
Table A-1. Model Equation Parameters.
Variable Definition
Unit
Source
P
DT
Pn
V
Gt
G
IE
It
Brake engine horsepower
Horsepower loss in the drive train
Horsepower loss from accessory
operations
Engine rotational speed
Vehicle speed
Power transmission efficiency
Transmission gear ratio
Final drive gear ratio
Effective moment of inertia
rotational moment of inertia of the
wheels and axles
Drive shaft rotational moment of inertia
Engine rotational moment of inertia
Transmission rotational moment of
inertia
bhp
A Axle-horsepower available for tractive ahp
work
hp
hp
rpm
ft/s
ft.lbf.s2
ft.lbf»s2
ft»lbf»s2
P = — -
E 5757
or P = P + P + P
E A DT a
N
A X
E
1 min
Gt xGdA60sec,
Ihp
550
ft-lbf
sec
Parameterized estimation 5%~12%
Parameterized estimation
Field measurement
V= 2w.
1 min
60 sec;
field measurement
Engine data books or typical value
Engine data books or typical value
Engine data books or typical value
ft»lb»s2
ft»lbf»s2 Parameterized estimation
f x G2d)x(l
Parameterized estimation
Parameterized estimation
Parameterized estimation
O
Q.
m
3
o
3
O
Q.
2.
5'
(Q
D)
1
O
m
£ >
C -D
Q) -D
? 0
O 3
3 Q.
"0 X'
^ >
fi>
3
0
ff
continued
-------
Table A-1. Model Equation Parameters (concluded).
Variable Definition Unit
Source
M,
Eff
m
a
FD
Fn
00
W
g
e
p
cd
Af
Effective mass
Drive wheel radius
Tractive force at the drive wheel
Total vehicle mass
Vehicle acceleration
Rolling resistance force
Gravitational force
Aerodynamic drag force
Rolling Resistance coefficient
Total vehicle weight
Acceleration of gravity
Inclination angle of the road
Air density
Atmospheric pressure
Atmospheric temperature
Drag coefficient
Vehicle frontal area
Effective vehicle velocity
lbm
ft
lbf
lbm
ft/s2
lbf
lbf
lbf
lbf
ft/s2
degrees
Ibf.s2/ft4
in.
mercury
°F
ft2
ft/s
Field measurement
FT = m x a + FR + Fw + FD
Field measurement
Field measurement
FR = Cr x mx gx cos(6>)
Fw = Wxsm(0)
FD = [|J x Q x Af x Ve2
Typical road surface coefficients
concrete, wet asphalt, asphalt, hot asphalt
W' =mxg
32.2 ft/s2
Field measurement or road construction database
- 0.00236l
-
519
V29.92A460+ TJ
Field measurement
Field measurement
Typical Cd for a flat top tractor (0.99) & high roof sleeper (0.60)
Field measurement
vw
Head wind velocity
ft/s
Field measurement
-------
Emissions Modeling Framework
Appendix B
Potential Data Sources for the Heavy-Duty Vehicle
Modal Emission Modeling Framework
V: vehicle speed (ft/s or m/s)
a: vehicle acceleration (ft/s2 or m/s2)
Method
Description
Suitability
Note
Car chip
VASCAR
Non-contact
speed sensor
Calibrated
Speedometer
Loop detector
Video detector
Radar and laser
gun
Racelogic
Velocity Box
An electronic device that is mounted
in-vehicle to record information such as
vehicle speed as a function of time
An in-car speed measuring computer that
records time taken to cover a distance,
thereby allowing average speed to be
calculated.
A sensor that is mounted on a vehicle,
pointing at the ground to measure the
speed of the vehicle relative to the
ground.
This allows one to measure vehicle
speeds for evidential purposes when
following a vehicle.
An inductive that is embedded in the
ground to collect traffic data such as
volume, speed, occupancy, etc.
Video cameras that is mounted along
roadside to collect traffic data such as
volume, speed, occupancy, etc.
A handheld device to measure vehicle
speed or distance.
Non Contact speed and distance
measurement using GPS (called VBOX)
which measures the speed, position,
acceleration figures, braking distances of
amoving vehicle.
Vehicle-specific
Vehicle-specific
Vehicle-specific
Vehicle-specific
Location-specific
Location-specific
Need to mount on a vehicle and
have someone to run the test
vehicle.
Need to mount on a vehicle and
have someone to run the test
vehicle.
Mainly for police use
Need to mount on a vehicle and
have someone to run the test
vehicle. More info:
http ://www. gmheng. com/pdf/an 1001 .pdf
(accessed August 2005)
For police use.
Automatic data and widely
available at traffic management
centers.
Time-consuming and
labor-consuming.
Location-specific Time-consuming and
Time-specific
Vehicle-specific
labor-consuming.
Resource-consuming.
More info:
http://www.m-techautomotive.co
.uk/vbox/VBox Index.htm
(accessed August 2005)
69
-------
Heavy-Duty Diesel Vehicle Model
Vw: Head wind velocity(ft/s)
Atmospheric winds vary in intensity throughout the
United States, with typical mean values of 10-20 mph, and
gusty winds to 50 and 60 mph. The atmospheric wind will
be random in direction with respect to the vehicle
direction of travel.
Method
Description
Suitability
Note
Anemometer
Air Velocity
Flow Sensor
Weather Vanes
Wind Socks
Weather station
Mounted on the roof of a vehicle to
measure the wind that results from
vehicle speed and direction as well as the
wind speed and direction.
Inserted into a duct or pipe through an
access hole to measure air velocity.
Weather vanes are one of the oldest of all
weather instruments, working by
swinging around in the wind to show
which direction it is blowing from.
Wind Socks show visual indication of the
wind.
It may have collected such information
systematically.
Vehicle-specific
Not clear
Wind direction
Wind direction
Wind speed
Need to mount on a vehicle and
have someone to run the test
vehicle.
More info:
http://sensors-transducers.globalspec.com/Sp
ecSearch/Suppliers?Comp=289 (accessed
August 2005)
http://www.rcn27.dial.pipex.com
/cloudsrus/measure wind.html
(accessed August 2005)
http://www.rcn27.dial.pipex.com
/cloudsrus/measure wind.html
(accessed August 2005)
Details will be on "Online
Resources of Climatic Database"
part
70
-------
Emissions Modeling Framework
£ Roadway Grade Angle
Numerous methods exist that can be employed to
determine roadway geometric characteristics such as the
grade and super-elevation. Depending on the purpose,
location, and available resources, these methods span from
conventional land surveying techniques to advanced
technologies such as photogrammetry and digital terrain
models. Several factors are important in selecting any one
method, and these include cost, time, work-crew safety,
and the desired accuracy (dissertation of H. Ikwut-Ukwa).
Method
Description
Note
Leveling Survey
Grade Gauge
Vangarde 505
Remote Sensing
Roadway design
blueprint
Contour map
This method relies on the determination of
relative elevations between points along the
road to determine the longitudinal and
lateral slopes; these are translated into
roadway grade and banking, respectively.
A reading is taken by simply placing the
gauge on the roadway where the slope is to
be measured, adjusting the arm that gives
the reading.
An infrared electronic distance measurer
(EDM) and a theodolite that allows
measurements to targets on the road surface
from a static, remote location.
Remote Sensing data are collected from
high altitude satellites, such as the
LAND SAT, or from high altitude aerial
photography.
May have grade information labeled on the
blueprint.
GIS centers may provide such electronic
maps and use AutoCAD to read contour
lines.
Relative elevations with leveling survey are
typically accurate to a hundredth of a foot.
Since surveyors need to be physically present
on the road, the use of this method sometimes
involves some restricted traffic operation; on
high speed roads such as freeways this is either
unsafe or impractical.
Surveyors need to be physically present on the
road. It is impractical to apply this method on
freeways.
This system provides data to an accuracy of
two millimeters. Though most of the work is
done from the vehicle, the system still requires
conventional survey to establish controls. The
system does not perform well on new asphalt
and on wet pavement because of the light
absorbing/scattering effect of these surfaces. It
is also very costly and time intensive.
The grade data obtained are inaccurate since
most existing Digital Elevation Models (DEM)
data have very low resolution.
The final construction details of a road may
differ significantly from the original plans
because of unanticipated conditions in the
field.
71
-------
Heavy-Duty Diesel Vehicle Model
W: Total vehicle weight (lbf)
The weight data can be measured from the field.
Method
Description
Note
Wight-in-motion
(WIM) Site
Weight Stations
National Truck
Survey
Permanent WIM locations are selected to monitor the
weights experienced by bridges and specific roads.
Weigh stations are located along
Interstate highways, usually near the border between
states. All trucks are required by law to be weighed at the
weigh station during its hours of operation.
Currently, there are four major national truck travel data
sources available which feature heavy-duty vehicle
characteristics: the Truck Inventory and Use Survey, the
Commodity Flow Survey, the Nationwide Truck Activity
and Commodity Survey, and the National Truck Trip
Information Survey.
Regional The two most common types of truck surveys are trip
Commercial Vehicle diaries and roadside surveys.
Surveys
Bending plate The device typically consists of a weigh pad attached to a
technology metal frame installed into the travel lane. A vehicle passes
over the metal frame causing it to slightly "bend." Strain
gauge weighing elements measure the strain on the metal
plate induced by the vehicle passing over it. This yields a
weight based on wheel/axle loads on each of two scales
installed in a lane. The device also is used to obtain
classification and speed data.
Source http://ntl.bts.gov/DOCS/arizona_report.html (accessed August 2005).
Generally weigh stations are
open only during the day,
depending on the nature of the
specific WIM equipment at a
particular station.
The Bureau of the Census
conducts the Truck Inventory
and Use Survey (TIUS) every
five years as part of the Census
of Transportation.
rjtot: the total efficiency of power trans-
mission
This number is determined from experiment. Typically,
this value(s) may be determined by testing a truck
engine(s) using a chassis dynamometer. If a chassis
dynamometer is unavailable for the testing, then 80% to
85% can be used as the default value.
Gt: the transmission gear ratio
The gear ratios in the transmission for 1st through nth gear
vary by vehicle make, model, and model year. For
example, a heavy-duty, "deep low" 5-speed manual
transmission, as used in a 2500 Series pickup, has the gear
ratio of 5.61:1, 3.04:1, 1.67:1, 1:1, and 0.75:1. The
transmission gear ratio for many models can be found
online: http ://www. vibrate software .com/
html help/html/Diagnosis/Transmission Gear Ratios
main.htm . Also, the diesel truck index includes this
information for many models. A library of drive train
technologies will be assembled over the coming months.
Gd : the differential, or the final drive,
gear ratio
Differential gear ratio determines the number of times the
drive shaft (or pinion) will rotate for each turn of the
wheels (or ring gear). Gear ratio is calculated by dividing
the number of teeth on the ring gear by the number of
teeth on the pinion gear. The higher the number, the lower
the ratio. Larger, heavier vehicles tend toward the higher
72
-------
Emissions Modeling Framework
numeric ratios in the differential.
Af: Vehicle frontal area
The frontal area is the projected area of the vehicle relative
to its direction of travel and is expressed as square feet. It
is used to determine aerodynamic drag losses on the
vehicle. The diesel truck index includes this information
for many models.
Cr: Rolling resistance coefficient
Rolling resistance is a measure of the amount of resistance
that is generated as a tire, which is deformed at the contact
to the ground, rolls on the road surface. This deformation,
which is a function of tire size and type, pavement type,
vehicle weight, and vehicle speed, can create rolling
resistance. Rolling resistance increases with increasing
softness of the road surface. The rolling friction
coefficient gives the force of friction needed to maintain
the uniform motion when it is multiplied by the normal
force between two bodies rolling with each other. This
coefficient is determined from experiment. The typical
value for truck ranges from 0.006 to 0.01.
Surfaces
Coefficient of Friction3
Rolling Friction
Kinetic Friction
Low-rolling-resistance car tire on dry
pavement
Ordinary car tire on dry pavement
Truck tire on dry pavement
Train wheel on steel track
0.006-0.01
0.015
0.006-0.01
0.001
0.8
0.8
0.8
0.1
1 Source: http://www.school-for-champions.com/science/frictionrolling.htm (accessed August 2005).
Typical Coefficient Values"
Surface
venicie lype
Concrete Medium Hard
Passenger cars
Heavy Trucks
Tractors
0.015
0.012
0.02
0.08
0.06
0.04
Sand
0.30
0.25
0.20
1 Source: Gillespie (1992)
Typical Coefficient of Rolling Resistance
Road Surface Cr
Pneumatic tires on:
Concrete asphalt 0.015
Rolled coarse gravel 0.02
Tarmacadam 0.025
Earth 0.05
Farmland 0.1-0.35
Wheel on rail 0.001-0.002
73
-------
Heavy-Duty Diesel Vehicle Model
Cd: drag coefficient
The drag coefficient is determined from experiment.
Typical aerodynamic drag coefficients provided by Ford
Motor Company are shown in Figure B-l. Drag
coefficient for heavy-duty vehicle varies, but a value of
0.99 is commonly used for flat top tractor and 0.60 for
high roof sleeper.
p = 0.00236
_
519
29.92 A 460 +
Parachute
Flat plate
CSqussr*?)
Flat top
tractor
High roo*
sleeper
m*-*m,
Cone (6uJ!
He mi sphe re=~-(>.'->-
Thunderbird
(ios4 rom;
Cone(30°)
Sphere
Airfoil
135
1.17
0,99s
0.60*
0.51
0.41
o
0.34
0 10
0.05
Drag Coefficient of Various Shapes
'Source: Ford Motor Co. "National Research
Council of Canada, **NASA)
Figure B-1. Typical Aerodynamic
Drag Coefficients.
p: air density
Air density is used to calculate the aerodynamic drag. If
this number is not available directly, it can be calculated
from atmosphere pressure and temperature. The function
is
where, Pr is the atmospheric pressure in inches of mercury
(Hg) and
Tr is the atmospheric temperature in degrees
Fahrenheit
At standard conditions (59 °F and 29.92 in. Hg), the
density is 0.00236 Ibrsec2/ft4.
Online air density calculators can provide air density by
altitude, temperature, altimeter setting, and dew point
information from a climatic database at either
http://wahiduddin.net/calc/calc da.htm or
http://www.denysschen.com/catalogue/densitv.asp (both
accessed August 2005).
• NCDC's Weather and Climate Resources:
° Get/View Online Surface Data
http://www.ncdc.noaa.gov/ol/climate/climatedata.htmltfSURFACE
(accessed August 2005),
° Hourly (Temperature, Precipitation, Winds,
Pressure, Etc, from 1997-present, sorted by
station). Fee will be charged!
° Daily (Temperature, Precipitation, Winds,
Pressure, Snow, Etc, CD-rom 1948-present),
° Monthly (Temperature, Precipitation, Pressure,
Etc, from 1800-1996),
° Modeled (1900 - present, monthly temperature
and precipitation),
• Climate Monitoring and Diagnostic Laboratory:
ftp ://ftp .cmdl .noaa. gov/met/ The directories contain
hourly average observations of air temperature,
station pressure, and surface wind direction and
speed at the four NOAA/CMDL observatories,
Point Barrow, Alaska (brw), Mauna Loa
Observatory, Hawaii (mlo), American Samoa
Observatory (smo), and Amundsen-Scott, South
Pole Observatory, Antarctica (spo).
• Climate Prediction Center (CPC) data:
http://www.cpc.ncep.noaa.gov/data/ (accessed
August 2005):
° Selected Historical Data,
° Weekly/Monthly Degree Days
http://www.cpc.ncep.noaa.gov/products/analysi
s monitoring/cdus/pastdata/degree days/
(accessed August 2005); contains degree days
data for the country for 3 weeks prior to current
74
-------
Emissions Modeling Framework
date and 3 months prior to current date,
° Weekly/Monthly Precip/Temp Tables
http: //www .cpc .ncep .noaa.gov/products/analvsi
s monitoring/cdus/pastdata/prcp temp/
(accessed August 2005); contains precipitation
and temp-erature data of the country for most
recent 3 weeks and most recent 3 months,
• NOAA Server:
Access to NOAA data and information:
http://www.esdim.noaa.gov/noaaserver-bin/NOAAServer/
Searchable interface to NOAA's data holdings.
Users can download and plot data. It is a place
where you can search database based on keywords,
time range, geographical coverage, database
searched, and search criteria. Search results are
given in forms like Description, Preview, Obtain,
and Ordering info.
Drive Wheel Radius
Rw: the outer radius of the drive wheel (inches)
rw. the inner radius of the drive wheel (inches)
The radius data can be gotten from the tire itself directly.
There is some small print on the tire's sidewall to specify
the design information, and the radius of the tire can be
calculated from them. For example, Figure B-2 shows a "P
215/65 R 15" tire. The 215 is the width of the tire in
millimeters measured from sidewall to sidewall. The 65 is
called "aspect ratio" and is used to tell height of the tire
from the bead to the top of the tread. This number is
described as a percentage of tire width. That means the
height is 215 x 65% = 139.75mm (5.59in). The 15 is the
rim diameter in inches to specify the wheel rim diameter
the tire is designed for. So the tire diameter is 2 x 5.59in
+15 inches = 26.18 in (654.5 mm). After these calcu-
lations, the outer radius of the drive wheel is 13.09 in, the
inner radius of the drive wheel is 7.5 in (half of the rim
diameter).
Ratio of height to width
(aspect ratio)
Width ot fire
in millimeters
Passenger
car tin
Radial, , Diameter of wheel in inches
Load index &
speed symbol
U.S. DOT safety
standard code
Max. cold inflation
& toad linvl
Treadwear traction and
temperature grades
Tire ply composition
and materials used
Figure B-2. Tire Detail Information.
Moments of Inertia
The diesel truck index includes tire information for many
models.
ID: mass moment of inertia of drive train
IE: mass moment of inertia of engine
IR: mass moment of inertia of wheel
These moment of inertia data can be used to calculate the
rational inertia coefficient (e). Generally, these data can be
obtained from laboratory experiments.
75
-------
Heavy-Duty Diesel Vehicle Model
Appendix C
Estimating Heavy-Duty Vehicle Miles Traveled
Heavy-Duty Vehicle Activity Data
Sources
Publicly available transportation databases managed by
Federal and State agencies were used to estimate heavy-
duty VMT within the 20-county Atlanta region. To
estimate HDV2b VMT, 2-axle, 4-tire vehicle VMT
percentages for road types from the Highway Statistics
Series (HSS) of FHWA were used. To distinguish light-
duty trucks (LDTs—gross vehicle weight ranges from
6,001 to 8,500 Ibs) and HDV2b (gross vehicle weight
ranges from 8,501 to 10,000 Ibs) from the 2-axle, 4-tire
VMT, Georgia statewide LOT and HDV2b VMT was
obtained from the Vehicle Inventory and Use Survey
(VIUS) of U.S. Census Bureau. Truck percent, segment
length, and annual average daily traffic (AADT) from
Georgia Department of Transportation (GDOT) Highway
Performance Monitoring System (HPMS) were used to
estimate classes HDV3 to 8b VMT. The Georgia Tech
HDV/BUS database developed in 2003 was used in
addition to Federal and State databases to separate aggre-
gated truck VMT from HPMS into EPA HDV classes.
Highway Statistics Series
The annually published HSS provides highway vehicle
activity information such as statewide annual total VMT
by road type and by FHWA truck class, which is classified
with the number of axles and truck-trailer combinations
(FHWA, 2001). Until 1999, the HSS had provided other
2-axle, 4-tire vehicle VMT percentages, which are the
mixture of LDT and HDV2b VMT, for each road type
except collectors and locals (FHWA, 2000). However,
since 2000, the HSS has not provided the other 2-axle,
4-tire vehicle VMT percentages by road type. Therefore,
the Class Two vehicle VMT percentages from pre-2000
HSS were used in this study. From the 1993 to 1999 HSS,
seven year average other 2-axle, 4-tire vehicle VMT
percentages, which are statistically significant means at
5% significant level, were used for each road type in this
study. Table C-l shows the average other 2-axle, 4-tire
vehicle VMT percentages by road type in Georgia state-
wide.
Table C-1. Georgia Statewide Other 2-Axle, 4-Tire
VMT Percentages.
Road Types
Average 2-Axle, 4-
Tire VMT Per-
centages
1
2
6
11
12
14
16
Rural interstates
Other principal rural ar-
terial roads
Minor rural arterial roads
Urban interstates
Other urban freeways
and expressways
Principal urban arterial
roads
Minor urban arterial
roads
20.5
13.3
15.2
26.1
25.9
23.0
21.6
Due to differences in HDV (or Truck) definitions between
EPA and FHWA, HPMS databases do notprovide HDV2b
VMT. These other 2-axle, 4-tire VMT percentages can be
also used to calculate total other 2-axle, 4-tire VMT from
total VMT from HPMS databases. Then, total other 2-axle,
4-tire VMT should be separated into each LDT VMT and
HDV2b VMT because the 2-axle, 4-tire vehicle VMT is
the mixture of LDT and HDV2b VMT.
Vehicle Inventory and Use Survey
A VIUS conducted by U.S. Census Bureau in 2002
provides statewide 2-axle, 6-tire HDV VMT with vehicle
GVWR, which can direct the conversion of surveyed
HDV VMT into EPA HDV class VMT (FHWA, 2004).
Because the VIUS does not provide countywide HDV
VMT, HDV VMT fractions from statewide HDV VMT
were used to separate the observed HDV VMT in the
Georgia Tech HDV database in 2003 into EPA HDV
classes 3 to 8, which correspond to HDV class X1 (see the
section, Georgia Tech HDV/BUS Database, below) from
the observed HDV VMT (Table C-2).
76
-------
Emissions Modeling Framework
Table C-2. Georgia Statewide HDV VMT Fractions
for the Class X1 Conversion.
GVW Ranges X
(Ibs) Class
10,001-14,000
14,001-16,000
16,001-19,500
"VI R
19,501-26,000
26,001-33,000
33,001-60,000
EPA HDV
Classes
HDV3
HDV4
HDV5
HDV6
HDV7
HDVSa
VMT
Fraction
0.2577
0.1824
0.0773
0.2726
0.1861
0.0239
Highway Performance Monitoring System
HPMS databases provide roadway segment lengths,
AADT, and truck percentages from which total VMT and
truck VMT can be calculated for each road type. However,
truck percentages do not count 2-axle, 4-tire HDV because
the definition of "truck" encompasses vehicles more than
or equal to 2-axle, 6-tire. That means that EPA HDV2b
VMT may not be included in the truck VMT estimated
from HPMS database. Using 2-axle, 4-tire vehicle VMT
percentages from HSS and the LDT/HDV2b VMT ratio
from VIUS, HDV2b VMT can be estimated with total
vehicle VMT from HPMS databases.
Georgia Tech HDV/BUS Database
HDV volumes were observed on a freeway and arterial
roadway network within the 21-county Atlanta region.
The highway network was composed of 90 freeway and
202 major arterial roadway segments falling within one
mile of the region's major warehouses and truck stops.
The 292 roadway segments were aggregated into 59
segment groups by roadway geometry similarity (contigu-
ous interchanges, roadway merges, and separations). The
goal was to combine similar traffic activity within a
segment group, so that one segment from each of the 59
segment groups could be randomly selected as represen-
tative of the group. HDV volumes collected on a selected
segment were then used for all segments in a segment
group. HDV volumes for the selected segments were
counted at sites using a visual HDV classification scheme,
which employed four HDV classes according to engine
horsepower and vehicle weight similarities (Ahanotu,
1999), for consecutive 2 hours. Not only were HDV
volumes counted, but school bus and other bus volumes
were also counted at the selected sites. Since HDV
volumes for each segment were observed only for 2 hours,
a scale-up method was used to scale up 2-hour HDV
volumes to 24-hour HDV volumes with representative
24-hour HDV volume profiles for each Ahanotu HDV
class observed on two freeway segments (1-285 and 1-20)
and one arterial segment (US-41) for consecutive 24 hours
on a weekday. HDV 24-hour volume profiles with each
Ahanotu HDV class on 1-285, 1-20, and US-41 were
scaled up 2-hour HDV volumes observed on the segments
of freeways and arterials. After the scale-up, HDV VMT
for each segment and each Ahanotu HDV class were
estimated. However, the Ahanotu HDV classes can not be
directly converted into EPA HDV classes, so that a
method(s) was needed for HDV class conversion.
To convert observed HDV volumes by Ahanotu HDV
class into EPA HDV classes, anew HDV visual classifica-
tion scheme—which is a hybrid HDV visual classification
scheme between FHWA and EPA HDV classification
schemes with FHWA truck weight limitations (FHWA,
1994; FHWA, 2002) and EPA GVWR—was developed in
2003 (Yoon, et al, 2004a). The new HDV classification
scheme (called the X-scheme) has three HDV classes from
the modification of the Ahanotu HDV classification
scheme. The three HDV classes are 2-axle HDVs (XI),
3-axle HDVs (X2), and more than 3-axle HDVs (X3)
(Figure C-3). The XI class corresponds to the Ahanotu
HDV class A, which is 2-axle, single unit HDVs, and
corresponds to EPA HDV3 to HDV7. Observed HDV
volumes of the X1 class can be apportioned to EPA HDV3
to HDV7 by multiplying their VMT fractions obtained
from 2002 VIUS. The X2 class corresponds to parts of
Ahanotu HDV class B (3-axle, single unit) and C (3-axle,
two units), which correspond to FHWA HDV classes 6 or
8, and directly mapped into the EPA HDVSa. The X3
X2
X3
HDV2b, HDV3,
HDV4, HDVS,
HDV6, HDV7
HDVSa
HDVSb
•' ~ £k'-!*"~
• fJttSStHgyf ~y^ *tp*yy 1 TIT *r « *"
V~
Figure C-1. Typical X-Scheme HDV Class
Examples.
77
-------
Heavy-Duty Diesel Vehicle Model
class corresponds to the part of Ahanotu HDV class B
(4-axle, single unit) and C (4-axle, 2-unit) and all of the
class D. The X3 class corresponds to FHWA classes 7 and
9 to 13, and directly maps into the EPA HDVSb.
Georgia Tech also developed a school bus activity data-
base, which is a part of the Georgia Tech HDV database,
through letter/telephone survey in 2003 within a 21-
county Atlanta region. The school bus database includes
the number of school buses and daily miles for each bus.
With the number of school buses and daily miles, school
bus VMT was calculated for each county. It was found
that total school bus VMT from the database was more
than two times higher than the observed school bus VMT
on the roadway network. This apparently is due to the fact
that school buses operate mostly on minor arterial and
local roads, while the roadway network database includes
only a small fraction of minor arterial roads and no local
roads. In essence, the roadway network database does not
reflect the real-world road network. Therefore, observed
school bus VMT on the roadway network should be
corrected with school bus VMT from the database. School
bus VMT correction will be conducted after HDV VMT
estimation on and off the roadway network because the
correction should be applied to only minor arterial and
local roads.
Heavy-Duty Vehicle VMT Estimation
All data sources discussed in the Heavy-Duty Vehicle
Activity Data Sources section are integrated to estimate
HDV VMT by road and by EPA HDV class within a
20-county Atlanta region. The overall process of HDV
VMT estimation with databases is described in Figure C-4.
Heavy-duty vehicle VMT from 2003 Georgia Tech HDV
database was assigned to each roadway segment to
identity road types through GIS analysis techniques.
Roadway segments on the freeway/arterial network (not
including collectors and locals) were assigned each
corresponding FHWA road types. HDV VMT with the
Ahanotu HDV classification scheme (Ahanotu, 1999) for
each road type was translated into EPA HDV class VMT
through an intermediate conversion via the X- scheme. For
the translation of the XI HDV VMT into EPA HDV
classes 3 to 7, HDV VMT fractions from VIUS (2002)
were applied. After VMT translation into EPA HDV
classes, HDV VMT fractions with EPA classes were
calculated and applied to apportion truck VMT estimated
from the GDOT HPMS database for each road type within
the 20-county Atlanta region.
The reason for using HDV VMT fractions from the
Georgia Tech HDV database instead of using national
default vehicle class adjustment factors (U.S. EPA, 2002b
and U.S. EPA, 1998) is to avoid underestimating VMT,
especially for HDVSa and HDVSb classes. Yoon et al.
(2004b) studied HHDDV travel patterns at seven truck
stops in and around the 21-county Atlanta region bound-
ary and found that over 50% of HHDDVs did not make
2002 Vehicle
Inventory and Use
Survey
'&
Mean Annual Miles
per vehicle for
HDV3-8A Classes
Registered HDV3-
8A Classes by Fuel
Type & Age
2003 Georgia Tech
School/Urban Bus
Database
2003 Georgia Highway
Performance
Monitoring System
HDV VMT by Facility Type
& HDV Class
Apply VMT Fractions for HDV3-SB, <ooIBia, & IHxtnBus
Figure C-2. Overall Process of HDV VMT Estimation.
78
-------
Emissions Modeling Framework
Table C-3. EPA HDV VMT Fractions for Road Types.
EPA HDV VMT Fractions
Koaa
Type
Freeway
Arterial
Local
HDV3
0.058
0.046
0.034
HDV4
0.041
0.032
0.024
HDV5
0.018
0.014
0.010
HDV6
0.062
0.048
0.036
HDV7
0.042
0.033
0.025
HDVSa
0.101
0.165
0.130
HDVSb
0.651
0.610
0.487
School
Bus
0.009
0.025
0.231
Other
Bus
0.018
0.028
0.022
Table C-4. HDV and Bus VMT Fractions within the 20-County Atlanta Region.
EPA HDV VMT Fractions
Koaa
Type
1
2
3
HDV2
b
0.170
0.242
0.339
HDV3
0.048
0.035
0.022
HDV4
0.034
0.024
0.016
HDV5
0.015
0.010
0.007
HDV6
0.051
0.037
0.024
HDV7
0.035
0.025
0.016
HDV8
a
0.083
0.125
0.086
HDV8
b
0.540
0.462
0.322
School
Bus
0.008
0.019
0.153
Other
Bus
0.015
0.021
0.015
any stops within the region but passed through. That
means that over 50% of HHDDVs may not be registered
within the region. If VMT fractions are generated with
registration data, HHDDV VMT may be severely underes-
timated while lighter HDV VMT may be highly overesti-
mated. Table C-3 shows EPA HDV VMT fractions
obtained from the Georgia Tech HDV database for road
types within the 20- county Atlanta region.
From total HPMS-estimated VMT, 2-axle, 4-tire VMT
was calculated from the average 2-axle, 4-tire vehicle
VMT percentages in the HSS. Then, the average 2-axle,
4-tire vehicle VMT percentages were divided by the
LDT/HDV2b VMT ratio to obtain HDV2b VMT for each
road type. From the HPMS database, total truck VMT was
also calculated using AADT and a truck percentage for
each link. After the estimation of HDV2b VMT, HDV
VMT fractions were generated for each road type within
the 20-county Atlanta region (Table C-4).
Total HDV VMT within the 20-county Atlanta region can
be downsized into each county HDV VMT through the
GIS spatial analysis. Because HPMS database was built
with unit segment information—which includes the road
typeand length and truck percent—county level HDV
VMT can be generated through database management
techniques. In addition, the Georgia Tech school bus
database was also built by city and county bases, and
therefore, school bus VMT can be generated for each
county.
Application in Road Load-Based Emis-
sions Modeling
Estimated HDV VMT by road type and by EPA HDV
class can be directly used in modal activity-based emis-
sions models for the regional on-road mobile source emis-
sions inventory development. In modal activity-based
emissions models, HDV VMT will be associated with
emissions rates in grams per brake-horsepower, tractive
horsepower, road length, and the fraction of a road
grade-length matrix at given conditions.
where E is the emissions in grams per day,
/' is the heavy-duty vehicle class,
/is the road type,
j is the speed in the speed-acceleration matrix,
k is the acceleration in the speed-acceleration
matrix,
/ is the engine model year,
79
-------
Heavy-Duty Diesel Vehicle Model
P is the tractive power in brake-horsepower-hour,
AFF is the acceleration frequency fraction in the
speed-acceleration matrix,
ER is the emissions rate in grams per brake-horse-
power-hour,
LMis the total lane length in miles, and
VMTis the total vehicle miles traveled in miles per
day.
80
-------
Emissions Modeling Framework
Appendix D
Emission Testing Contacts
West Virginia University
Engine and Emissions Research Laboratory
Greg Thompson, Research Assistant Professor
Phone: 304-293-2419. Gregorv.Thompson(fl),mail.wvu.edu
Transportable Emissions Laboratory
Nigel N. Clark, Professor, Department of Mechanical and
Aerospace Engineering,
Phone: 304-293-3111. nclark@,wvu.edu
University of California Riverside
College of Engineering-Center for Environmental Re-
search and Technology (CE-CERT)
Transportation Systems and Vehicle Technology Research
Laboratory
Mattew Earth, Associate Professor,
Phone: 909-781-5782, barth@,cert.ucr.edu
Mobile Heavy-duty Diesel Emissions Laboratory
Lisa Arth, Special Programs, CE-CERT,
Phone: 909-781-5665, lisa(g),cert.ucr.edu
U.S. Department of Energy National
Renewable Energy Laboratory (NREL)
Renewable Fuels and Lubricants Research Laboratory
(ReFUEL Laboratory):
Margo Melendez
Phone: 303-275-4479, margo melendez@nrel.gov
Douglas Lawson, Principal Scientist,
Phone: 303-275-4429, doug lawson@inrel.gov
U.S. EPA National Vehicle and Fuel
Emissions Laboratory (NVFEL)
Terry Newell
Phone: (734) 214-4462, newell.terry@epa.gov
California Air Resources Board Heavy
Duty Emissions Testing Laboratory
(ARB-HDETL)
Alberto Ayala, Research Division, Air Resources Board,
California Environmental Protection Agency,
Phone: (916) 327-2952, aavala@arb.ca.gov
North Carolina State University
Department of Civil Engineering
Christopher Frey
Phone: 919-515-1155. frev@eos.ncsu.edu
81
-------
Heavy-Duty Diesel Vehicle Model
Appendix E
Heavy-Duty Vehicle Emission Rate
Data Sources and Applications
MOBILE series models, which were developed by U.S.
EPA and are used in 49 states for regulatory purposes,
estimate HDV emissions rates based on certified engine
dynamometer test results. Certified engine dynamometer
emissions expressed in grams per brake-horsepower-hour
will be converted into emissions in grams per mile with
conversion factors for each HDV classes from HDV2b to
HDVSb. Emissions rates from engine dynamometer tests
will be corrected with various correction factors to repre-
sent real-world emissions rates. In addition, the EMFAC-
2000 emissions model that is only used in California
estimates HDV emissions rates based on chassis dyna-
mometer test results. Chassis dynamometer emissions
rates expressed in grams per mile can be directly used to
estimate emissions rates without using conversion factors.
However, emissions rates from both dynamometer test
results may not correctly represent horsepower corre-
sponding emissions rates on the road. That is because
horsepower requirement varies by road grades, weight,
speed, and acceleration according to real road conditions.
The load-based (required horsepower) HDV emissions
model framework, which is under development by Geor-
gia Institute of Technology (Georgia Tech), has the same
model framework concept used for MOVES. To link
required horsepower at a specific road and vehicle operat-
ing conditions to emissions rates in grams per brake-
horsepower-hour, grams per axle-horsepower-hour, or
grams per mile, the research team has reviewed available
data from various HDV emissions test laboratories and
developed strategies on how to incorporate the data to
Georgia Tech load-based HDV emissions model frame-
work.
Emission Rate Data Available for
Analysis
NVFEL (National Vehicle and Fuel Emissions
Laboratory, EPA)
NVFEL provides engine dynamometer test emissions rates
(zero mile emission levels plus deterioration rates) in
grams per brake-horsepower-hour, which were tested with
the Federal test procedure (FTP) HDV transient cycle
(U.S. EPA, 2004). Emissions rates provided from this
laboratory are used in MOBILES and MOBILE6 mobile
source models. Before incorporating the engine dynamom-
eter test emissions rates with required horsepower from
the proposed emissions model framework, horsepower
losses through the drive train and differential should be
considered. Because the brake-horsepower from engine
dynamometer tests indicates the net horsepower available
at the engine crankshaft, horsepower losses through the
drive train and differential should be excluded from the
net available horsepower from engine dynamometer tests.
For use in MOBILE6, Tables E-l to E-3 show heavy-duty
engine emission rates by model year group for HC, NOX,
and CO. All the emissions rate data are available from
Lindhjem and Jackson, 1999.
Table E-1. Heavy-Duty Vehicle HC Emission Rates (Grams per Brake-Horsepower-Hour) for Use in MOBILE6.
Light Medium Heavy
Year
1989
1990
ZMLa
Gd
0.62
0.35
De
0.64
0.52
Det@
G
0.023
0.023
110kb
D
0.002
0.001
ZML
G
0.62
0.35
D
0.66
0.52
Det@
G
0.023
0.023
1 185kc
D
0.002
0.001
ZML
G
0.62
0.35
D
0.47
0.52
Det @ 290kc
G D
0.023 0.001
0.023 0.000
continued
82
-------
Emissions Modeling Framework
1991- 0.33 0.47 0.021
1994- 0.33 0.26 0.021
1998- 0.33 0.26 0.021
2004+ 0.33 0.26 0.021
0.001
0.001
0.001
0.001
a ZML = zero mile level.
b Det @ = deterioration rate at 1 10k mi.
c The useful life of all heavy-duty gasoline engines is
c G = gasoline engine.
d D = diesel engine.
Table E-2. Heavy-Duty Vehicle
MOBILES.
Light
Year ZMLa Det @
Gd De G
1989 4.96 4.34 0.044
1990 3.61 4.85 0.026
1991- 3.24 4.38 0.038
1994- 3.24 4.08 0.038
1998- 2.59 3.26 0.038
2004+ 2.59 1.61 0.038
0.33
0.33
0.33
0.33
11 Ok mi.
NOX Emission
0.40
0.31
0.31
0.31
Rates
0
0
0
0
.021
.021
.021
.021
(Grams
0.
0.
0.
0.
per
001
001
001
001
0.33
0.33
0.33
0.33
0.30
0.22
0.22
0.22
0.021
0.021
0.021
0.021
Brake-Horsepower-Hour) for
Medium
110kb
D
0.002
0.011
0.003
0.001
0.001
0.001
a ZML = zero mile level.
b Det @ = deterioration rate at 1 10k mi.
c The useful life of all heavy-duty gasoline engines is
c G = gasoline engine.
d D = diesel engine.
ZML Det (c
G
4.96
3.61
3.24
3.24
2.59
2.59
11 Ok mi.
D
6.43
4.85
4.53
4.61
3.69
1.84
0
0
0
0
0
0
G
.044
.026
.038
.038
.038
.038
£ 185kc
0.
0.
0.
0.
0.
0.
D
009
006
007
001
001
001
0.000
0.001
0.001
0.001
Use in
Heavy
ZML
G
4.96
3.61
3.24
3.24
2.59
2.59
D
6.28
4.85
4.56
4.61
3.68
1.84
Det@
G
0.044
0.026
0.038
0.038
0.038
0.038
290kc
D
0.010
0.004
0.004
0.003
0.003
0.003
Table E-3. Heavy-Duty Vehicle CO Emission Rates (Grams per Brake-Horsepower-Hour) for Use in MOBILES.
Light
Year ZMLa Det @
Gd De G
1989 13.84 1.21 0.246
1990 6.98 1.81 0.213
1991- 7.10 0.40 0.255
1994- 7.10 1.19 0.255
1998- 7.10 1.19 0.255
2004+ 7.10 1.19 0.255
Medium
110kb
D
0.022
0.012
0.004
0.003
0.003
0.003
ZML Det (c
G
13.84
6.98
7.10
7.10
7.10
7.10
D
1.70
1.81
1.26
0.85
0.85
0.85
0
0
0
0
0
0
G
.246
.213
.255
.255
.255
.255
£ 185kc
0.
0.
0.
0.
0.
0.
D
009
006
007
001
001
001
Heavy
ZML
G
13.84
6.98
7.10
7.10
7.10
7.10
D
1.34
1.81
1.82
1.07
1.07
1.07
Det@
G
0.246
0.213
0.255
0.255
0.255
0.255
290kc
D
0.010
0.004
0.004
0.003
0.003
0.003
a ZML = zero mile level.
b Det @ = deterioration rate at 110k mi.
c The useful life of all heavy-duty gasoline engines is 11 Ok mi.
c G = gasoline engine.
d D = diesel engine.
83
-------
Heavy-Duty Diesel Vehicle Model
Table E-4.
Vehicle
Model
Year
1989
1994
1996
Body of file
1996
Heavy Heavy-Duty Diesel Vehicle Emission Rates from NREL.
Engine
Model
Year
1989
1994
1996
not shown
1996
Test
Cycle
CBDa
wvub
5 min
for brevity
5 min
Max
bhp
350
C
330
330
GVWR
(Ibs)
46,000
68,000
80,000
80,000
Emission
Test Year
1995
1995
1997
1998
Odometer
Reading
496,232
11,300
132,700
204,200
Avg. Emission
(g/mi)
HC NOX
1.37 12.7
2.02 33.2
35.2
31.8
Rate
CO
27.9
3.0
2.0
1.7
a CBD = central business district.
b WVU = West Virginia University.
c Max bhp unknown.
NREL (National Renewable Energy Labora-
tory, DOE)
National renewable energy laboratory provides chassis
dynamometer test average emissions rates in grams per
mile (NREL, 2004). The chassis dynamometer average
emissions rates come with detailed vehicle and engine type
information such as vehicle model year, engine model
year, maximum brake horsepower, test fuel type, vehicle
type, odometer readings, GVWRs, test year, average
emissions rates, and test cycles. All information provided
from the NREL is only for HHDVs. Table E-4 shows a
sample of information available from NREL. Before
incorporating NREL emissions rates with required horse-
power from the proposed model framework, average
emissions rates in grams per mile can be divided by
maximum brake horsepower and multiplied by test length
in miles (given by test cycle). Because vehicles were tested
multiple times with various test cycles, it is possible to
conduct meaningful statistical analysis for emissions rates.
However, the maximum horsepower values provided by
NREL are not clearly described. The values may be
interpreted to denote rated (maximum ) horsepower or
actual horsepower used in the tests. Measured horsepower
(actual) or a "book value" may have been used in the tests.
Depending on which horsepower value may have been
used in the tests and subsequent calculations, the converted
emissions rates in grams per brake-horsepower may be
greater or smaller than actual emissions rates. Table E-4 is
an example of the emission rates that may be obtained
from the data.
EERC (Engine and Emissions Research
Center, WVU)
Chassis dynamometer test results by WVU-EERC are
available through various research reports; those are CRC
Project No. E-55/E-59 (Gautam and Clark, 2003), CRC
Project No. E-55-3 (Clark and Gautam, 2004), and HDDV
emission test data from New York State Department of
Environmental Conservation (EEA, 2000). Available data
are in various units such as grams per brake-horsepower-
hour, grams per axle-horsepower-hour, and grams per
mile. If raw data from WVU-EERC are provided, the
research team can easily convert them into emissions rates
in grams per axle-horsepower-hour units before incorpo-
rating them with the required horsepower in the proposed
model framework. Table E-5 shows the example emis-
sions rates with axle horsepower from CRC Project No.
E-55/E-59.
Table E-5. Emissions Rates from CRC Project
No. E-55/E-59.
Test ahp-
Truck ID „ , ,
Cycle hr
E55CRC-1
E55CRC-1
E55CRC-1 avg
D
D
D
14.88
15.37
15.12
Emission Rates
(g/ahp-hr)
CO
4.72
5.60
5.16
N0xa
13.38
12.81
13.10
N0xb
13.29
13.22
13.25
HC
0.09
0.09
0.09
continued
84
-------
Emissions Modeling Framework
E55CRC-2 D 16.74
E55CRC-2 D 17.04
E55CRC-2 avg D 16.89
1.29 6.38 6.54 0.28
1.37 6.25 6.29 0.26
1.33 6.32 6.41 0.27
a Analyzer number 1.
b Analyzer number 2.
New York State Department of Environmental Conser-
vation and Energy (NYSDEC) test data are also available
from EMFAC2000 documentation (CARB, 2000), which
provides engine model year, gross vehicle weight ratings,
actual vehicle weight, and emissions rates (Table E-6).
Table E-6. NYSDE Test Data Used in EMFAC2000.
Model GVWR
Year (Ibs)
1985
1989
Body
1997
26
33
of file
33
,000
,000
Test
Weight
(Ibs)
18,
23,
200
100
Odometer
(miles)
21,
66,
600
300
Emission Rates
(g/mi)
HC
0
0
.15
.63
CO
2
6
.33
.09
NOX
12.10
18.80
not shown for brevity
,000
23,
100
3500
0
.08
4
.93
16.60
CIFER (Colorado Institute for Fuels and High
Altitude Engine Research)
CIFER at the Colorado School of Mines provided chassis
dynamometer test results from the northern front range
study and opacity inspection (Colorado IFHAER, 2004).
The test results were also used to develop emission rates in
EMFAC2000. In EMFAC2000 documentation, CIFER
data shows engine model year, gross vehicle weight
ratings, actual vehicle weight, odometer, start status (hot or
cold), and emissions rates (Table E-7).
Table E-7. CIFER Test Data Found in EMFAC2000
Document.
Model
Year
1990
1993
1993
1993
GVWR
(Ibs)
33,000
25,500
25,500
25,500
Test
Weight
(Ibs)
23,667
18,049
18,049
18,049
Odometer
(miles)
142,242
122,406
122,406
122,406
Start
(hot
or
cold)
hot
cold
hot
hot
Emission
(g/mi
HC
0.26
1.24
0.56
0.62
NOX
15.41
14.97
13.82
13.39
Rates
)
CO
4.93
18.41
CAEC (Cleaire Advanced Emission Controls,
LLC)
CAEC provided the research team with a set of HDV
chassis dynamometer test results. The test results show
second-by-second emissions rates and engine parameters
during chassis dynamometer testing (CAEC, 2004).
However, the test results do not provide second-by-second
brake (or axle) horsepower, instead providing only rated
horsepower at a given RPM.
EMFAC2000
Emissions rates in EMFAC2000 model are based on
chassis dynamometer test results unlike MOBILE6, which
is based on engine dynamometer test results. To develop
emissions rates in the EMFAC2000 model, California Air
Resources Board (CARB) used data tested by NYSDEC,
WVU-EERC, and CIFER. Through statistical analysis,
emissions rates in grams per mile for diesel light, medium,
and heavy HDVs were estimated. Table E-8 shows the
diesel HHDV average emissions rates (zero mile emis-
sions plus deterioration rates) used in EMFAC2000.
Table E-8. Diesel HHDV Emissions Rates (grams
per mile).
Year
1987-
1991-
1994-
1998
1999-
2003
2004+
HC
ZMLa
0
0
0
0
0
0
0
.34
.28
.19
.18
.18
.14
.14
10kb
0
0
0
0
0
0
0
.009
.009
.016
.014
.009
.003
.003
CO
ZML
2
1
0
0
.48
.74
.84
.63
0.63
1
1
.01
.01
10k
0
0
0
0
0
0
0
.065
.056
.068
.049
.031
.023
.023
NO
ZML
16
15
19
23
.79
.97
.06
.01
13.36
6
6
.68
.68
X
10k
0.015
0.030
0.042
0.037
0.013
0.007
0.007
a ZML = zero mile level (engine has zero miles).
b engine deteriorated to 10k miles.
NCSU (North Carolina State University)
Recently NCSU measured instantaneous medium HDV
engine activity and emissions using an onboard portable
monitoring system while the vehicle was running on the
road. They measured NOX, PM, CO, and CO2 emissions
with fuel use, vehicle speed, and location. However, they
did not measure horsepower from the vehicle.
85
-------
Heavy-Duty Diesel Vehicle Model
Application in the Load-Based HDDV
Modal Emission Modeling Framework
Emissions rates from chassis dynamometer test laborato-
ries described above could be used for the development of
the load-based HDV emissions model framework. Chassis
dynamometer test results from WVU-EERC would be the
most reliable because they have tested an extensive
number of HDVs (light HDVs for NYSDEC, medium
HDVs for CRC Project No. E-55-3, and heavy HDVs for
CRC Project No. E-55/E-59) and provided emissions rates
in grams per axle-horse-power-hour, test methods, and axle
horsepower in axle-horsepower-hour. These data can be
directly incorporated with required horsepower for vehicle
activity.
LER = ER x P/3600
where, LER is the load-based emissions rate in grams per
second,
ER is the chassis dynamometer test result in grams
per axle-horsepower-hour,
P is the tractive power for vehicle activity in axle-
horsepower per second, and
3600 is the conversion factor from hours to sec-
onds
For further detailed load-based emissions modeling with
WVU-EERC data, more detail data than is available in the
published reports (raw data) should be provided.
However, emissions rates from the chassis dynamometer
test do not extensively incorporated with second-by-
second vehicle activities involved in road grades and
off-test cycle acceleration and speed. That means that the
emissions rates from the chassis dynamometer test may
not represent second-by-second emissions rates although
they can be expressed in grams per second after
multiplying required horsepower. In this case, emissions
rates and axle (wheel) horsepower measured at the same
time by CE-CERT would be more real-world representing
emissions data. However, CE-CERT does not provide test
route elevation incorporated with second-by-second
emissions rates and horsepower. In addition, their data
may not statistically significant because they do not
provide enough data.
86
-------
Emissions Modeling Framework
Appendix F
Modeling Approaches On-road Heavy-Duty Diesel Vehicle
Oxides of Nitrogen and Particulate Matter Emissions:
PARTS, MOBILES, EMFAC7G, and EMFAC2000
This appendix discusses the basis of oxides of nitrogen
(NOX) and diesel PM emission rate modeling from on-road
HDDVs in current models.
HDDVs are the major on-road NOX and diesel PM emis-
sions sources, which impacts high concentrations of
ground level ozone and fine particulate matter in the
atmosphere. To analyze the air quality impact of the
pollutants from HDDVs, national and state air quality
agencies develop on-road emissions inventories using
mathematical emission models such as PARTS,
MOBILE6, EMFAC7G, EMFAC2000, and later models.
PARTS, developed by U.S. EPA in 1995, was designed to
estimate particulates from tailpipes, tires, and brakes.
Since then, PARTS became the prototype of a PM emis-
sion model for other emission models. As PARTS descen-
dants in diesel PM emissions estimation, EMFAC7G and
MOBILE6 were developed by California DOT in 1997
and by U.S. EPA in 2001. In 2000, California DOT
developed a new generation emission model (EMFAC-
2000), which adopted anew concept to estimate emission
rates from on-road HDDVs. EMFAC2000 estimates
emission rates with chassis dynamometer test results,
whereas the other models predict the emission rates with
the engine dynamometer test results. Diesel PM emission
rates estimated with EMFAC2000 were significantly
different from its predecessor, EMFAC7G. In this report,
this diesel PM emission rate difference will not be dis-
cussed because that is beyond the objective. CARB
provides detailed information of the difference in diesel
PM emissions rates between the two models. Table F-l
shows the emissions models capable of diesel PM emis-
sions estimation.
MOBILE6, EMFAC7G, and EMFAC2000 can estimate
NOX emissions from on-road HDDVs. NOX emission
estimation with MOBILE6 and EMFAC7G are based on
engine dynamometer test results. From these engine dyna-
mometer tests, base NOX emission rates were determined
and then are corrected with series of correction factors.
Conversely, base NOX emission rates on EMFAC2000 are
determined from chassis dynamometer test results, which
were provided by U.S. EPA, NYSDEC, and WVU. The
base NOX emission rates are corrected with series of
correction factors as well. Like diesel PM emissions rates,
NOX emissions rates estimated with EMFAC2000 were
significantly different from EMFAC7G. This NOX emis-
sions rate difference will also not discussed in this report
for the same reason the diesel PM emission rate difference
is not discussed. CARB released enhanced EMFAC
models in 2001 and 2002. However, the concept for
estimating NOX emissions is same as with the EMFAC-
2000. Therefore, only EMFAC2000 will be discussed in
this report. Table F-2 shows the emission models capable
of NOX emission estimation.
Table F-1. PM Emissions Estimation—Primary Model Components.
Model Core Components
PARTS
BER3 determined by engine dynamometer test results
BER =y(model year, speed, speed cycle, number of tires
TotPM = ExhPM + BrakePM + TirePM
ExhPM = OCPMb + ECPMC + sulfate
CarbonPM (g/mi) = BER (g/bhp-hr) x CFd (bhp-hr/mi)
Idling ERe (g/hr)
continued
87
-------
Heavy-Duty Diesel Vehicle Model
EMFAC7G • BER determined by engine dynamometer test results
• BER =y(model year, speed, speed cycle, number of tires)
• BELf =^(ZML, DETR8)
• TotPM = ExhPM + BrakePM + TirePM
• ExhPM = OCPM + ECPM + Sulfate
• CarbonPM (g/mi) = BER (g/bhp-hr) x CF (bhp-hr/mi)
MOBILE6 • BEL determined by engine dynamometer test results
• BEL =y(model year, speed, speed cycle, number of tires)
• BEL =/ZML, DETR)
• TotPM = ExhPM + BrakePM + TirePM
• ExhPM = OCPM + ECPM + Sulfate
• CarbonPM (g/mi) = BER (g/bhp-hr) x CF (bhp-hr/mi)
EMFAC2000 • BEL (g/mi) determined by chassis dynamometer test results
• BEL =/model year, speed, temperature, off-cycle, etc.)
• BEL =/ZML, DETRf)
• ERh (g/mi) = BEL x Corrections (model year, speed, temperature, off-cycle, etc)
• Idling ER fe/hr)
a BER = basic emission rate.
b OCPM = organic carbon PM emissions.
c ECPM = elemental carbon PM emissions.
d CF = conversion factor.
e ER = emissions rate.
f BEL = diesel initial (baseline) PM emisions.
g DETR = deterioration rate..
Table F-2. NOX Emissions Estimation—Primary Model Components.
Model Core Components
EMFAC7G • BEL3 (g/bhp-hr) determined by engine dynamometer test results
• BEL =J[model year, speed, temperature, off-cycle, etc.)
• BEL =f(7ML, DETRb)
• EL° =y(sales fraction, horsepower fraction, BEL)
• ERd (g/mi) = EL (g/bhp-hr) x CFe (bhp-hr/mi)
MOBILE6 • BEL (g/bhp-hr) determined by engine dynamometer test results
• BEL =J[model year, speed, temperature, off-cycle, etc.)
• BEL =./(ZML, DETR)
• EL =y(diesel sales fraction, horsepower fraction, BEL)
• ER (g/mi) = EL (g/bhp-hr) x CF (bhp-hr/mi)
EMFAC2000 • BEL (g/mi) determined by chassis dynamometer test results
• BEL =J[model year, speed, temperature, off-cycle, etc.)
• BEL =f(7ML, DETR)
• ER (g/mi) = BEL x Corrections (model year, speed, temperature, off-cycle, etc)
• Idling ER fe/hr)
a BEL = diesel initial (baseline) PM emisions.
b DETR = deterioration rate.
c EL = average emission rate for each vehicle type.
d ER = emissions rate.
e CF = conversion factor.
-------
Emissions Modeling Framework
PARTS
PARTS emission model released by U.S. EPA in 1995 can
estimate diesel PM emission rates from on-road HDDVs.
Diesel PM emission rates, which are a function of vehicle
model year, speed, speed cycle (transient and steady), the
number of wheels, and so on, consist of carbon PM, direct
sulfate, brake-wear, and tire-wear emissions. Carbon PM
emissions include organic carbon and elemental carbon
emissions from HDDVs. Because initial (base) carbon
emissions are expressed in particulate grams per brake-
horsepower-hour, they are converted in particulate grams
per mile with the conversion factor in brake-horsepower-
hour per mile. The expressions of elemental and organic
carbon emissions follow.
x
FD
BSFCx
(F-l)
(F-2)
(F-3)
where ECPM is the elemental carbon PM emissions in
grams per mile,
BEL is the diesel initial (baseline) PM emissions in
grams per brake-horsepower-hour,
CF is the conversion factor in brake-horsepower-
hour per mile,
DSF is the direct sulfate emissions in grams per
mile,
FEC is the elemental carbon fraction of the diesel
exhaust emissions factor,
OCPM is the organic carbon PM emissions in
grams per mile,
FD is the fuel density in pounds per gallon,
BSFC is the brake specific fuel consumption in
pounds per brake-horsepower-hour,
FE is the fuel economy in miles per gallon,
m is model year, and
v is vehicle class.
Direct sulfate emissions are calculated with the assump-
tion that all sulfur in diesel fuel is exhausted as sulfate or
sulfate dioxide.
SUPMmv = ll.5xFDx SWPxDSCF/FEmv
(F-4)
where SUPM is the direct diesel sulfate emissions in
grams per mile,
SWP is the sulfur weight percent in diesel fuel, and
DSCF is the direct sulfur conversion percent to
sulfate.
From the equations F-1 to F-4, base exhaust emission rates
can be determined for each vehicle class and vehicle
model year. In PARTS, base exhaust diesel PM emissions
can be determined with the zero mile emissions rate and
deterioration rates, which were obtained from Federal Test
Procedure (FTP) test results.
/ \
^ \ ' m'V (F-5)
(CM2xDTR2)
^ ' m,v
where ExhPMis the exhaust base emissions rate in grams
per mile,
ZM is the zero mile emissions rate in grams per
mile,
CMl is the cumulative mileage less than useful life
(mileage),
DTRl is the deterioration rate at CMl5
CM2 is the cumulative miles at useful life (mile-
age) minus CMl5 and
DTR2 is the deterioration rate at CM2.
Then, the base exhaust emissions rates are weighted with
series of correction factors such as speed, temperature,
high emitter correction factors, and so on for each vehicle
type and vehicle model year. Brake-wear emissions rate is
uniformly applied to all vehicle classes.
BrakePM= 0.0128 x PSBRK
(F-6)
where BrakePM is the brake-wear PM emissions in
grams per mile and
PSBRK is the particle fraction to the particle size
cutoff.
Tire-wear emissions are a direct function of the average
number of wheels (tires) for the vehicle type.
TirePMv = 0.002 x PSBRK x ANWV (F-7)
where TirePMis the tire-wear PM emissions in grams per
mile and
ANWis the average number of wheels.
89
-------
In addition, PARTS displays diesel HDDV idle emissions
rates in grams per hour, which were collected from vehicle
manufacturers. Idle emissions rates, however, only varies
by vehicle year group, but not by vehicle type.
EMFAC7G
EMFAC7G (a model in MVEI7G package) estimates
diesel PM emissions from on-road HDDVs, using an
approach almost identical to PARTS, except for tire-wear
particle size cut-off fraction for PM10. The particle size
cut-off fraction used in EMFAC7G was 0.4 for PM10 out
of total suspended particulate (TSP), while PARTS used
1.0 for the particle size cut-off fraction.
NOX emissions from HDDVs in EMFAC7G can be
calculated with average emissions levels and correction
factors for each vehicle type and vehicle model year. The
concept to estimate NOX emissions with EMFAC7G is
same with MOBILES and MOBILE6.
MOBILE6
MOBILE6 also used the same estimation approach that
PARTS used to calculate diesel PM emissions. However,
the linear relationship between zero mile emissions rate
and mileage in MOBLE6 may differ from that in PARTS
because MOBILE6 uses some modified correction factors
from original PARTS correction factors.
EL =
I (SalesVtyr
xHP
v,tr
(F-8)
where ELV is the average emissions rate for each vehicle
type,
Sales is the diesel sales fraction,
HP is the engine horsepower, and
EL is the base emissions rate for each vehicle
type and vehicle model year.
Then, the average emission rates in NOX grams per
brake-horsepower-hour can be converted by the conver-
sion factor expressed in the Equation F-3. In MOBILE6,
the fuel economy and the brake specific fuel consumption
are expressed with curve fit equations for vehicle model
year using Truck Inventory and Use Survey data and truck
manufacturers' brake specific fuel consumption test
results.
EMFAC2000
In diesel PM and NOX emissions estimation from on-road
HDDVs, EMFAC2000 uses a completely different ap-
proach from other emissions models. EMFAC2000 uses
chassis dynamometer test results to estimate base emis-
sions rate for diesel PM and NOX emissions, whereas other
models use engine dynamometer test results. Because
EMFAC2000 uses base emissions rate in pollutant grams
per mile, it does not need to use conversion factors to
change emission rate units. To develop diesel PM and
NOX curve fit equations for vehicle model year,
EMFAC2000 uses chassis dynamometer test results
(tested with urban dynamometer test schedule—UDDS)
collected from U.S. EPA,NYSDEC, andWVU. Then, the
base emissions rate can be weighted with series of correc-
tion factors. Because U.S. EPA, NYSDEC, and WVU
only provide the UDDS results of medium and heavy
HDDVs, EMFAC2000 uses the FTP test results done by
CE-CERT (UC Riverside) for light HDD Vs. EMFAC2000
also recalculate zero mile emissions rate and deterioration
rate to reflect tampering and maintenance component
effect on the collected data.
NOX emissions from HDDVs in MOBILE6 can be calcu- ZM =
lated with average emissions levels and correction factors
for each vehicle type and vehicle model year. Baseline
NOX emissions can be determined with Equation F-5. Base
emission rates for each vehicle model year can be aver-
aged among each vehicle type.
EL
2\
(F-9)
DR=(ER- ZM}/\Odometer/lOOOO) (F-10)
where ZMis the zero mile emissions rate,
ER is the average emissions rate of the chassis
dynamometer data,
£/! is the emissions impact prediction of the
Radian model (HDDV I/M study by Radian
Corporation) assuming the effect of both tamper-
ing and mal-maintenance,
EI2 is the emissions impact prediction of the
Radian model assuming the effect of only tamper-
ing, and
DR is the deterioration rate.
For the first time in EMFAC models, EMFAC2000
introduced idling emissions rates (grams per hour), which
are for HDDVs running with speeds of less than 5 mph
and a trip length of less than 5 miles.
90
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/R-05/090a
2.
3. RECIPIENTS ACCESSION NO.
4. TITLE AND SUBTITLE
Heavy-Duty Diesel Vehicle Modal Emission Model (HDDV-
MEM) Volume I: Modal Emission Modeling Framework
5. REPORT DATE
August 2005
6. PERFORMING ORGANIZATION CODE
7. AUTHORS
R. Guensler, S. Yoon, C. Feng, H. Li, J. Jun
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
School of Civil and Environmental Engineering
Georgia Institute of Technology
Atlanta, GA
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
4C-R022-NAEX
12. SPONSORING AGENCY NAME AND ADDRESS
U. S. EPA, Office of Research and Development
Air Pollution Prevention and Control Division
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final; 11/03-02/05
14. SPONSORING AGENCY CODE
EPA/600/13
15. SUPPLEMENTARY NOTES
The EPA Project Officer is E. Sue Kimbrough, Mail Drop E305-02, phone (919) 541-2612, e-mail:
kimbrough.sue@epa.gov
16. ABSTRACT
The report outlines research of a proposed heavy-duty Diesel vehicle modal emission modeling framework
(HDDV-MEMF) for heavy-duty diesel-powered trucks and buses. Although the heavy-duty vehicle modal
modules being developed under this research are different from the motor vehicle emissions simulator
(MOVES) model, the HDDV-MEMF modules should be compatible with MOVES. In the proposed
HDDV-MEMF, emissions from heavy-duty vehicles are predicted as a function of hours of on-road operation
at specific engine horsepower loads. Hence, the basic algorithms and matrix calculations in the new
heavy-duty diesel vehicle modeling framework should be transferable to MOVES. The specific
implementation approach employed by the research team to test the model in Atlanta is somewhat different
from other approaches in that an existing geographic information system (GIS) based modeling tool is being
adapted to the task. The new model implementation is similar in general structure to the previous modal
emission rate model known as the mobile assessment system for urban and regional evaluation
(MEASURE) model. This exploratory framework is designed to be applied to a variety of policy
assessments, including those aimed at reducing the emission rates from heavy-duty vehicles and those
designed to change the on-road operating characteristics to reduce emissions.
17.
KEYWORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Highway Transportation
Trucks
Buses (vehicles)
Emissions
Transportation Models
Pollution Control
Stationary Sources
13B
15E
13F
14G
12A
18. DISTRIBUTION STATEMENT
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
104
Release to Public
20. SECURITY CLASS (This Page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION IS OBSOLETE
forms/admin/techrpt.frm 7/8/99 pad
------- |