Peer Review of the Greenhouse Gas
Emissions Model (GEM) and
EPA's Response to Comments
Phase II
&EPA
United States
Environmental Protection
Agency
-------
Peer Review of the Greenhouse Gas
Emissions Model (GEM) and
EPA's Response to Comments
Phase II
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
Prepared for EPA by
Versar, Inc.
EPA Contract No. EP-C-12-045
Work Assignment No. 44
&EPA
United States
Environmental Protection
Agency
EPA-420-R-15-009
June 2015
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
TABLE OF CONTENTS
A. INTRODUCTION 1
B. CHARGE TO REVIEWERS 3
C. SUMMARY OF PEER REVIEWER COMMENTS BY QUESTION 6
GENERAL IMPRESSIONS 7
CHARGE QUESTION 1 7
CHARGE QUESTION 2 8
CHARGE QUESTION 3 9
CHARGE QUESTION 4 10
CHARGE QUESTION 5 10
D. PEER REVIEWER COMMENTS BY CHARGE QUESTION 12
Table 1. General Impressions 13
Table 2. Charge Question 1 18
Table3. Charge Question 2 27
Table 4. Charge Question 3 54
Table 5. Charge Question 4 57
Table 6. Charge Question 5 60
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"65
Table 8. Specific Observations on Electronic Model Entitled, "GEM Tool" 76
E. INDIVIDUAL PEER REVIEWER COMMENTS 82
Peer Reviewer # 1 83
Peer Reviewer # 2 92
Peer Reviewer # 3 108
Peer Reviewer # 4 114
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
A. INTRODUCTION
EPA, the National Highway Traffic Safety Administration (NHTSA), and the California Air
Resources Board (CARB), in looking to reduce greenhouse gas emissions (GHG) and to improve
fuel efficiency in medium- (MD) and heavy-duty (HD) vehicles, are considering recognizing the
efficiency of various powertrain technologies within the context of any new full vehicle emission
standard(s). For this option, it becomes critical to develop methods that assess the expected real
world performance of those technologies, including vehicle engine, transmission and axle
technologies.
Enhancements have also been made to their HD vehicle simulation software tool, GEM
(Greenhouse Gas Emission Model). At present, GEM is used by vehicle manufacturers to certify
the expected GHG emissions of their products. With the enhancements, GEM could potentially
have the ability to model a majority of the advanced technologies being incorporated into these
vehicles and their engines and that are being recognized by engine and chassis dynamometer
emission testing today.
EPA and the other agencies consider the GEM tool as a principal support for the second round of
HD GHG emissions regulations which are under development at the present time in both
NHTSA and EPA. The model has undergone a formal peer review in an earlier iteration of the
GEM tool (Phase I) and this newest version of GEM (Phase II) is the subject of this peer review.
EPA is looking to assure the regulated community of the high quality of the agencies' predictive
tool and that the proposed structure (and overall development process) of the GEM model results
in a tool that is simple, accurate and well-suited for the diversity of vehicles to which it may be
applied. The purpose of the requested peer review is for EPA to receive written comments from
experts on the concepts and methodologies upon which GEM relies and whether or not the
model can be expected to execute these algorithms correctly.
The purpose of the requested letter review is for EPA to receive written comments from
individual experts on GEM Phase II tool and supporting documentation ("Vehicle Simulation
Model").
Versar selected four senior scientists with expertise/experience in the following areas to serve as
peer reviewers. The reviewers are familiar with the use of models to characterize vehicle
simulations/operations; specifically, model design and model code and logic. Additionally,
reviewers have expertise in one or more of the following areas:
• vehicle operations and analysis, including the physical process of generating and
controlling vehicle emissions;
• linkages between mobile source emission modeling and transportation modeling and
planning; and
• application of current mobile source emissions models, w.r.t, heavy-duty vehicles, for
analysis for regulatory purposes and/or policy evaluation, e.g., HD GHG Notice of
Proposed Rulemaking.
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Reviewers:
Christopher M. Atkinson, Sc.D.
Atkinson, LLC
Morgantown, WV 26508
Nigel N. Clark, Ph.D.
West Virginia University
Morgantown, WV 26506-6203
Oscar F. Delgado-Neira, Ph.D.
The International Council on Clean Transportation
Washington, DC, 20005
Ashok Nedungadi, Ph.D., PE
Future Is Now Consulting
San Antonio, TX 78256
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
B. CHARGE TO REVIEWERS
EPA developed a forward-looking MATLAB/Simulink-based tool, the Greenhouse Gas
Emissions Model or GEM, for Class 2b through Class 8 vehicle compliance in 2011. At present,
this model is being used by all medium- and heavy-duty vehicle manufacturers for their vehicle
certifications. In order to meet EPA's upcoming Phase 2 GHG rulemaking requirements,
recognizing most of the vehicle technologies that are measured by both engine and chassis
dynamometer testing, the original GEM tool (versionl) has been considerably enhanced.
Specifically, the following key technical features have been implemented into GEM:
• Engine controller, including engine fuel cut-off algorithm during braking and
deceleration, and more stable engine idle speed controller;
• Transmission model, which includes upgraded manual transmission, newly developed
automatic transmission and automated manual transmission; and
• Upgraded driver model, with distance-based driver.
EPA has comprehensively validated its GEM model against over 130 vehicle variants, where all
tests, including vehicle chassis and powertrain dynamometers, are conducted at Southwest
Research Institute in San Antonio, TX. The summarized comparisons can be seen in the figure
below. As can be seen, good agreements between simulation and testing data were obtained, with
a few exceptions. For those outliers, it was found that the chassis roller may not be using friction
force to correctly represent the truck running on the road, thus causing the chassis dynamometer
testing to underestimate fuel consumption.
^ 4 6 8 10 12 14 16
Experimental Tests (MPG)
Figure 1. Comparisons between simulations and tests over 130 vehicles under tests.
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
It should be pointed out that CO2 reduction stringency standards in all certification categories are
derived from GEM simulation runs by first selecting a range set of vehicle parameters and
engine fuel maps. All to-be-certified vehicles from vehicle manufacturers use their own vehicle
parameters together with some of the agencies pre-defined parameters to run GEM, and then
their results are compared to what the agencies proposed. Since the comparisons are done in a
relative base to see if the vehicles can meet standards, it is hoped that this kind of accuracy
shown below would be adequate for use of certification purpose.
Please note that GEM includes a pull-down menu mechanism by which it is able to account for
those technologies that GEM is unable to model at the present time in development (due to either
time constraints or lack of experimental data needed to validate the model). Unlike a model/tool
used for research and development simulations where the model can be continuously upgraded
and developed, GEM as a compliance tool is constrained by legislative authorities which makes
both upgrading the tool and delivering the updated version to users difficult.
Charge Questions
EPA's vehicle simulation model, GEM, was created to serve as the primary tool to certify Class
7/8 combination tractors and trailers and Classes 2b - 8 vocational vehicles in meeting EPA's
and NHTSA's proposed vehicle GHG emission levels and fuel efficiency requirements. As both
agencies' proposed compliance tool, GEM needed the following modeling attributes:
1) Models a wide array of MD and HD vehicles over various drive cycles;
2) Built with open source code (provides transparency in the model's operation);
3) Available free to any user with minimal, or no, prior modeling experience;
4) Provides both optional and preset elements at input; and
5) Managed by both EPA and NHTSA for compliance purposes.
In general, the purpose of this peer review is for EPA to receive written comments on the
concepts and methodologies upon which the model relies and whether or not the model can be
expected to execute these algorithms correctly. In making comments on the model, please
distinguish between recommendations for clearly defined improvements that can be readily made
based on data or literature reasonably available and improvements that are more exploratory or
dependent on information not readily available. Any comment(s) should be sufficiently clear and
detailed to allow a thorough understanding by the Agencies or other parties familiar with the
model.
1. Please comment on EPA's overall approach to the stated purpose of the model (meet agencies'
compliance requirements) and whether the particular attributes found in the resulting model
embodies that purpose. Were there critical results or issues that were not discussed or addressed
by the GEM tool or its component sections?
2. Please comment on the appropriateness and completeness of the contents of the overall model
structure and its individual systems and their component models (i.e., using the
MATLAB/Simulink version), if applicable, and considering the following:
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
a) Elements in each system used to describe different vehicle categories;
b) Performance of each component model including the reviewer's assessment of the
underlying equations and/or physical principles coded into that component;
c) Input and output structures and how they interact with the model to obtain the
expected result, i.e., fuel consumption and CO2 over the given driving cycles; and
d) Default values used for the input file, as shown in "Vehicle Simulation Model"
document.
3. When using the standard of good engineering judgment, is the program execution optimized
by the chosen methodologies?
4. Please comment on the clarity, completeness and accuracy of the intended output/results (CO2
emissions or fuel efficiency output file).
5. In your opinion, are there any procedures or observations that would have added to the quality
of the GEM tool? Any recommendations for specific improvements to the functioning of the
outputs of the model?
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
C. SUMMARY OF PEER REVIEWER COMMENTS BY QUESTION
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Summary of Peer Review Comments on EPA's Heavy-Duty Greenhouse Gas Emission
Model (GEM): Phase II and Supporting Documentation
GENERAL IMPRESSIONS
Overall, the reviewers found EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM):
Phase II and supporting documentation to be well structured, articulate, and a valuable resource
for the EPA to represent the relative efficiency and emissions of heavy-duty vehicles. Reviewers
commented that the tool will be very useful for compliance purposes since it is a reasonable
compromise between accounting for all technology differences and advances, while remaining
simple to execute. A number of improvements in the model and supporting documentation were
suggested by the reviewers.
One reviewer commented that the overall dynamic performance of the model prediction is
difficult to judge because the results present a single-valued, integrated snap-shot of the model
prediction, and the time-based instantaneous fuel efficiency and carbon dioxide emissions
predictions are not available for review.
Another reviewer commented that the material does not address quantitatively the experimental
error that is likely in these types of comparisons, and measurement errors associated with rolling
resistance and drag are not included in the data comparison. Overall agreement with
experimental data (as vehicle fuel economy) does not validate each component model to the
same degree, and this should be acknowledged more clearly.
Additional suggestions were on the supporting documentation. One reviewer noted that there is
an overall lack of key detail on certain technical features, such as descriptions of features, how
they were developed, and quantitative results in several areas.
CHARGE QUESTION 1
Please comment on EPA's overall approach to the stated purpose of the model (meet agencies'
compliance requirements) and whether the particular attributes found in the resulting model
embodies that purpose. Were there critical results or issues that were not discussed or
addressed by the GEM tool or its component sections?
In general, the reviewers found that the EPA's overall approach to meet the agency's compliance
requirements is valid using the GEM Phase II model. The reviewers commented that a successful
approach must yield well-defined, unambiguous results for manufacturers and agencies and also
incorporate real world driving dynamics. All of the reviewers acknowledged that the input
functions of the model were substantially more beneficial than the original GEM Phase I model.
For example, adding fuel maps to the simulation by adding actual maps and drivetrain
parameters makes the results more realistic, allows the model to capture the effects of matching
engine and driveline, and ideally promotes right sizing of the engines to application.
One limitation a reviewer noted was that the model is oriented toward diesel engines. Naturally
aspirated gasoline, boosted gasoline, and natural gas engines are likely to increase in the next
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
five years and may warrant separate and careful consideration because their characteristics,
torque curves, and efficiency maps differ substantially from the diesel engine properties.
A few reviewers noted some limitations to the model that were unaddressed by GEM Phase II
model. For instance, it is possible for the vehicle not to meet the driving cycle as a result of
excessive grade or weight or other issues with the transmission/engine - the unaddressed issue in
this regard is a feedback alert to the user during the time instances when the vehicle is
significantly slowed down or does not meet the desired driving cycle. Another limitation is that
GEM Phase II model does not address thermal characteristics of the engine cooling system or the
heat rejection of the transmission fluids - these thermal issues affect the operational duty-cycle
of the engine fan, which will affect fuel economy.
One reviewer noted that in order for the public to have confidence in the regulatory program that
is built on a mix of engine- and vehicle-model-specific inputs and modeled GEM outputs, the
underlying data should be presented in full in downloadable data files (e.g., in Excel) as in
other EPA regulations.
Additional issues and limitations are addressed in the reviewers' individual responses to the
question.
CHARGE QUESTION 2
Please comment on the appropriateness and completeness of the contents of the overall model
structure and its individual systems and their component models (i.e., using the
MATLAB/Simulink version), if applicable, and considering the following:
a) Elements in each system used to describe different vehicle categories;
Overall, the model structure and its systems are appropriate and mostly complete for the
designated purpose of the model. The specific selection of engines and transmissions chosen will
cover a large portion of the current heavy-duty vehicle fleet. The modular structure and the
hierarchical arrangement of modules to mimic a real vehicle system makes the integration of
additional modules and capabilities easier to implement. One reviewer commented that GEM
Phase II cannot predict the benefits of learning algorithms, look-ahead strategies, and intelligent
vehicle systems for the optimization of powertrain efficiency on specific routes, but
acknowledged that emerging approaches would take great effort to configure.
One reviewer provided detailed comments on specific issues regarding model structure,
individual systems, and default values, including: fixed payloads, drop-down technologies driver
subsystem, transmission subsystem, engine fueling maps, modeling of idle cycle, trailers, and
accessories.
b) Performance of each component model including the reviewer's assessment of the
underlying equations and/or physical principles coded into that component;
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
The reviewers generally found the underlying equations and/or physical principles to be
sufficient to characterize real world driving situations. For example, one reviewer commented
that the distance-based approach was an important step to represent real life, but also commented
that accelerations and grade of cycles must also represent real life. Other suggestions for
improvements were also made. A couple reviewers commented on the steady-state maps and on
the transient adjustment factor (TAP). For example, one reviewer noted that the use of an engine
model steady-state map with a generic TAP does not encourage manufacturers to improve
transient fuel efficiency, and provided two possible solutions. Another reviewer commented that
in the case of nominally steady operation, the use of a steady-state fueling map is well-justified,
but the quasi-steady assumption required to allow the extension of the use of such a map to
transient operation requires additional justification. The reviewer also noted that the use of a
single TAF for a specific engine or powertrain configuration has the potential to cause prediction
inconsistencies. Comments were also made for other components, such as axle efficiency. One
reviewer noted that it can be dangerous to employ the overall fuel economy computation to
compare two approaches to modeling a single component, particularly if that component
represents a small loss.
c) Input and output structures and how they interact with the model to obtain the expected
result, i.e.,fuel consumption and CO2 over the given driving cycles;
One reviewer commented that the format for the input and output structures follows good coding
standards, making the input/output structures easy to use. One concern is whether there are any
other parameters or fitting techniques used to obtain the observed correlations that were not
listed in the documentation. Another reviewer commented, as guidance, that the manufacturers
will be obliged to use the executable version of GEM Phase II a large number of times to
compute an average value for compliance.
d) Default values used for the input file, as shown in "Vehicle Simulation Model" document.
Generally, the default data of GEM Phase II is complete and appropriate to execute a simulation
of heavy-duty vehicle powertrain over one of the drive cycles available in the default drive cycle
library. However, one reviewer thought it was not clear under what circumstances the user will
be able or allowed to make modifications in the final model implementation. For instance, it is
not clear that engine cooling fan loads have been adequately accounted for in the model as these
are typically not considered in the engine dynamometer testing from which engine fueling maps
are normally derived, and they can modify observed fueling rates by about 10% or more under
specific engine operating conditions.
CHARGE QUESTION 3
When using the standard of good engineering judgment, is the program execution optimized
by the chosen methodologies?
The reviewers generally agreed that the chosen methods and execution of the model shows
strong engineering judgment throughout. Two reviewers commented that it is difficult to
comment on the execution optimization of GEM-II. However, one of these reviewers stated that
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
the results, computational time, and outputs displayed indicate that the chosen methodologies are
suitable. The other reviewer stated that it appears that there are no S-functions within Simulink,
which would make the execution faster if there are any non-standard user defined functions.
Another reviewer commented that modeling accuracy would be raised if certain missing
components were considered, such as longitudinal slip of tires, rolling resistance during
crosswind correction, and the effects of yaw on drag.
CHARGE QUESTION 4
Please comment on the clarity, completeness and accuracy of the intended output/results (CO2
emissions or fuel efficiency output file).
Some of the reviewers provided suggestions for making the data reports more complete. Some of
these suggestions included adding (or clearly identifying) results for each drive cycle, adding
actual simulated speeds and a measure of deviation from the speed-distance, and summarizing
the results of the energy audit. One reviewer commented that users might be tempted to scale the
load-based results in an inappropriate fashion. Another reviewer commented that the results are
sufficiently comprehensive for the user who is not executing the program in MATLAB for
research and design purposes, and that for a manufacturer a very succinct output would be
sufficient.
CHARGE QUESTION 5
In your opinion, are there any procedures or observations that would have added to the quality
of the GEM tool? Any recommendations for specific improvements to the functioning of the
outputs of the model?
The reviewers provided a number of recommendations for improving the quality of the tool or
functioning of the model output. One reviewer suggested that GEM provide two different output
reports, including one in an aggregated format with information only relevant for compliance
purposes and one in a disaggregated format by cycle with other detailed information. In addition,
the output fields should provide a range of valid and acceptable values for parameters such as
number of shifts, ratio of number of shifts to number of gears in the transmission, distance
traveled, and ratio of actual time to target time. Another suggestion was to make consistent the
number of significant digits during each simulation, as the current presentation has a varying
number of significant digits which does not meet recommended practices in the presentation of
results and data. One reviewer also recommended adding language to remind users that if GEM
is used to compare two competing but very different technology packages, it may not have the
fidelity or granularity to evaluate which is better. Another recommendation was to provide a
quantitative discussion of choice of test weight for GEM. A further step suggested by one
reviewer would be to allow users to input their own cycles as is currently done with the VECTO
(Vehicle Energy Consumption calculation Tool) model in Europe. A number of other potential
model enhancements were recommended by another reviewer, such as including the ability to
create input data from a user provided spreadsheet, allowing users to select plots of key
component performance, having more detailed explanations of user provided data, providing
feedback during execution (i.e. percent complete), having pre-defined sample input data
10
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
available, including the ability to turn on the feed forward term for the driver model, and having
the ability to model accessory power draw in certain cases.
11
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
D. PEER REVIEWER COMMENTS BY CHARGE QUESTION
12
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 1. General Impressions
Reviewer Name
Reviewer Comment
EPA Response to Comment
Reviewer 1
The purpose of GEM in determining standards, providing a
compliance tool, and estimating real-world benefits is clearly
articulated. The proposed Phase II GEM tool has evolved
substantially from the Phase I version, particularly by allowing
additional user inputs that are necessary to acknowledge fuel
efficiency design improvements. As a compliance tool GEM
should seek to account for all technology advances and
differences, yet remain simple to execute and employ readily-
measured and well-defined input variables. The Phase II GEM
model has reached a reasonable compromise between these
conflicting goals, and this is made clear in the narrative. The
overall architecture is sound, the present component models are
appropriate to the task in nearly all cases, and they are clearly
described. The model yields credible results and credible responses
to input variable changes. The overall model predicts fuel
efficiency to within 5% of experimental measurements, and a clear
summary of these comparisons is presented. However, the material
does not address quantitatively the experimental error that is likely
in these comparisons, and measurement errors associated with
rolling resistance and drag are not included in the data comparison.
Overall agreement with experimental data (as vehicle fuel
economy) does not validate each component model to the same
degree, and this should be acknowledged more clearly. As one
example, overall fuel economy data are used to compare a
component model with a fixed efficiency value. The difference
overall was only 1.67%, but this represented more than a third of
the losses for that component. The move to a distance-based
strategy is justified and well-described, and represents a laudable
advance. Addition of several transmission models is presented, as
are thoughts on the addition of transient adjustment factors. Little
is said of the "powertrain variant architecture," although this may
Significant progress has been made since we
initiated the GEM Peer Review. The technical
research workshop held at Southwest Research
Institute on December 10-11, 2014 features the
progress done so far to the public. In this
workshop, both experimental errors and GEM
errors were addressed. The detailed progress
reports can be seen at
http://www.epa.gov/otaq/climate/regs-
heavy-duty.htm.
The supporting document, "Vehicle Simulation
Model", does provide partial validations on the
component level, such as the engine and
transmission, for selective cases. Due to the
large number of vehicle tests and GEM
simulations on over 130 vehicle variants, it
would be challenging to show every single case
validation at a component level. More
importantly, the certification is only conducted
at the final integrated and weighted fuel
efficiency and CO2 level, and therefore, the
focus of the model validations is on the overall
performance comparisons.
The supporting document, "Vehicle Simulation
Model", indicates that hybrid vehicles would
not be part of the certification with GEM, and
therefore, the powertrain described in this
report only talks about the conventional
powertrain system.
We are seeking comment in the NPRM
regarding the proposed transient correction
13
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 1. General Impressions
Reviewer Name
Reviewer Comment
EPA Response to Comment
prove important for integrated powertrains with or without a
hybrid component. Transient Adjustment Factors are not yet
finalized, and transient operation may warrant more than a single
correction factor approach.
factor of 1.05. We will finalize the transient
correction factor after receiving all comments
from stakeholders.
Reviewer 2
This document summarizes the findings of the review of the US
EPA's Heavy-Duty Greenhouse Gas Emission Model (Phase 2
GEM) and supporting documentation ("Vehicle Simulation
Model"). The tool will serve as the principal support for the second
round of Heavy-Duty GHG emissions regulations, which are under
development by NHTSA and EPA. The agencies are considering
recognizing the efficiency of various vehicle, engine, and
transmission technologies and they consider critical to develop
methods that assess the expected real world performance of those
technologies. The main purpose of this review is to evaluate how
well the developed model can serve as a regulatory and
compliance tool. The following represent my review of the tool
and accompanying report based on my experience in modeling
heavy-duty vehicles with full-vehicle simulation tools, as well as
from our assessment at the International Council on Clean
Transportation (ICCT) of other heavy-duty vehicle regulatory
models used around the world.
After reviewing the Matlab/Simulink model and the accompanying
report, my general impression is that the "Phase 2 GEM"
constitutes a valuable development effort by EPA to develop a
rigorous tool that represents the relative efficiency and emissions
of vehicles. The new modeling tools' comprehensiveness, quality
and amount of data inputs, and modeling structure reflect state-of-
the-art modeling techniques and accurately represent relative
efficiency differences of vehicles in real-world conditions.
The literature on vehicle simulation modeling is
abundant over the last few decades. Some of
the literature used in the supporting documents
provides some equations for certain
components.
The "Vehicle Simulation Model" document
was intended to represent a high level overview
of the purpose and scope of Phase 2 GEM for
regulatory purposes and therefore does not
provide detailed technical insight into the
model implementation. This document closely
resembles the Draft RIA Chapter 4 included
with the Notice of Proposed Rulemaking
(NPRM) for the HD Phase 2 program. Also the
final RIA Chapter 4 will add a few more
references to provide more insight into the
design and implementation of the model.
14
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 1. General Impressions
Reviewer Name
Reviewer Comment
EPA Response to Comment
The model architecture is clear and easy to follow and has
incorporated some key features that will enhance its overall
accuracy with respect to real world performance of technologies,
and will allow the model to capture fuel consumption reductions
from a broader range of technologies. Overall, the tool offers a
rigorous and comprehensive simulation accounting of both engine-
specific and full-vehicle effects in a manner that is suitable for the
regulatory compliance purposes as indicated. The model will be
capable of performing its intended purpose of reflecting
technology benefits for compliance purposes of most of the
technologies that the agencies are considering.
Some new vehicle modeling features are especially important,
namely the ability of the model to incorporate user-defined engine
fueling maps and driveline parameters, the development of
different transmission options, the enhanced transmission gear-
shifting strategy, the inclusion of a distance-based routes with road
grade, and the more comprehensive treatment of vocational truck
technologies. The accompanying testing effort that was undertaken
to validate the model is impressive and thorough, as capturing the
effect of combinations of technologies in such close agreement
with powertrain and chassis dynamometer testing is a difficult
task. The model development demonstrates a thorough
development process, and also shows a strong commitment to
transparently presenting the data and methodology that were
involved.
The comments below provide additional details, as well as some
suggestions that could also be considered by the agencies in the
final model development.
15
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 1. General Impressions
Reviewer Name
Reviewer Comment
EPA Response to Comment
Although the tool itself offers a suitable modeling platform, the
document that describes modeling approach could provide further
details in a number of areas. It appears as though the
documentation available for this peer review was at an early draft
stage. There is an overall lack of detail on key technical features
that are new in the model. Interested readers would gain from
better descriptions of such features, how they were developed, and
perhaps, more quantitative results in several areas. Also, the
quality of the report may be enhanced with more consistent use of
terminology and a reduction in the number of self-references.
Further details regarding areas where such documentation and
enhanced information would be helpful are described below.
Reviewer 3
The EPA Heavy-Duty Greenhouse Gas Emission Model (GEM)
Phase 2 documentation accurately represents the structure, format,
logic and algorithmic description of the model as presented. The
supporting documentation is for the most part, clear and self-
explanatory. The results produced by the GEM model appear to be
sound, although each set of results is presented as integrated fuel
efficiency and carbon dioxide emissions results. As such the
results present a single-valued, integrated snap-shot of the model
prediction, and the time-based instantaneous fuel efficiency and
carbon dioxide emissions predictions are not available for review.
The overall dynamic performance of the model prediction is thus
difficult to judge in the greater context of what is usually fully
transient vehicle operation. Furthermore, in the version reviewed,
the ability to vary input parameters and vehicle and drivetrain
attributes is limited to modifying input files and not through a
graphical user interface as described in the review instructions.
We do agree that addition of a graphic user
interface (GUI) could help the reviewers to
understand the code better. However, the
agencies do not intend to develop a GUI for
single GEM runs based on discussions with the
regulated stakeholders.
Instead of a GUI, the NPRM includes a macro-
enabled input file based on an Excel
spreadsheet (in addition to a standalone
executable) with some simple interactive
features (highlighting of out of range variables,
for example) and a single button to allow batch
processing of GEM runs.
16
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 1. General Impressions
Reviewer Name
Reviewer Comment
EPA Response to Comment
Reviewer 4
Accuracy of information:
The document provided, "Vehicle Simulation Model" provides a
good background on GEM-II, its differences from GEM-I, and the
Phase I certification process. Section 1.2, "Model Code
Description" describes the model components and sub-
components, in adequate detail for the user to understand the depth
and breadth of GEM-II. No underlying equations are provided.
The section on Model validation is an important section for GEM-
II. The extent of the validation and the comparisons with
dynamometer testing is impressive. This gives the reader
additional confidence in the results produced by GEM-II. The
validation process is well documented, concluding with the graph
that summarizes all the 130 vehicle validations performed. The
validation also includes graphs of component performance (engine
speed, engine fuel rate, and transmission gear number) as a
function of time
The document does not explain the variable target.veh_style. The
structure format used for the user input data is useful in collecting
all user-provided data.
A final conclusions section is missing in the document provided.
'target_veh_style' is an internal variable used
to track the regulatory sub category of the
simulation run and is not part of the user
interface.
A final conclusion section could be added into
the document, but it is not a critical element,
since the documentation was included in the
Draft RIA Chapter 4 and is more like a user
guide to help the user to understand the
structure of the program.
17
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
Reviewer 1
Throughout this review, except where different fuels are
mentioned, fuel efficiency improvement and GHG reduction are
considered to be synonymous.
Beyond the need for assurance of compliance is the need to reduce
fuel use and climate change emissions from heavy-duty vehicles in
revenue use. A successful EPA approach must be examined from
two perspectives. First, it is necessary that the approach yields
well-defined, unambiguous results to allow a manufacturer to
compare against a standard without the process being unreasonably
onerous. Second, it is necessary that that the vehicle attributes and
behaviors embodied in the certification process are substantially
representative of real world circumstances. These two necessities
are often in conflict, because the real world scenario is complex,
variable, and long in duration, whereas the standard must be
concise and precise. The conflict is far greater in the heavy duty
trucking arena than for light-duty vehicles or rail because truck
architectures vary widely, and are used in an even wider fashion.
Employing a model such as GEM to assure compliance provides
some relief in the conflict described above because the model can
be executed for a variety of activities and scenarios without
excessive cost of time and resources. However, as the model is
challenged to predict these varied (and emerging) scenarios with
fidelity, the model complexity rises. As a result, the empirical
tables or computational sub-models needed within the model grow
The agencies discuss the proposed regulatory
approach of separate engine and vehicle
standards in the preamble to the HD Phase 2
NPRM and how it may impact the criteria
pollutants. We will consider the comments
received during the comment period and make
any needed changes in the final rule.
18
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
in number and demand substantial engineering time to prepare and
verify. One might argue that "the more one models, the more one
measures," recognizing too that precise test protocols must
accompany each measurement.
Compromise is necessary between four fundamental needs:
1. Relevance of the model to real world truck operation.
Else the real world improvements will not match the
changes in standards
2. Accuracy of the model in predicting measured fuel
efficiency (and GHG production).
Else confidence in the model will be lost and the
compliance will become artificial
3. Accuracy of the model in predicting differential effects of
technology changes on fuel efficiency (and GHG
production).
Else the drivers for technology advancement will be lost
4. Control complexity and cost of modeling and compliance.
Else the cost transferred to the consumers will be
inappropriate
The overall new GEM approach shows awareness of these
necessities, and reaches a reasonable compromise. However, when
some parameters are fixed by the agency, manufacturers may be
discouraged from pursuing certain development opportunities in
the following way. Both future fuel pricing and future fuel
efficiency standards are unknown. If high fuel prices transcend the
19
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
standards, then manufacturers will pursue every cost-effective tool
to reduce fuel use. However, if fuel prices are low and are not the
driving force, and the GEM approach offers default values for
factors such as engine transient adjustment factors or transmission
efficiencies, some opportunities for real world reduction may be
left on the table. GEM cannot rely on economic drivers to address
technology advances that are not modeled.
GEM is a vehicle-based tool that is geared towards road-load
demands rather than engine-specific ("%load" and "%speed")
demands. Only the new distance-based cycles give a nod toward
engine power. Criteria pollutants are still characterized using a
paradigm based on engine rated output. Allowing different
measurement methodologies for efficiency and criteria pollutant
production will open a window for separate hardware and software
optimization for each test. This may demean the benefits of the
separate standards to some degree: in-use compliance for criteria
pollutant measurement will not close this gap because
measurement allowance for criteria pollutants reduces stringency
for criteria pollutants.
Reviewer 2
The proposed Phase 2 standards are predicated on the performance
of a broader range of technological improvements than Phase 1,
including changes to transmissions and better integration of
engines and transmissions, so a more comprehensive model is
required. The model in its current form will be capable of
performing its intended purpose of reflecting technology benefits
for compliance purposes of most of the technologies that the
In principal, we agree with the comments
that the input and output data should be
available to public. The agencies will
consider in a separate action whether the
GEM inputs and outputs submitted by
manufacturer to demonstrate compliance
with the Phase 2 standards are confidential
20
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
agencies are considering.
The model is enhanced in various aspects with respect to its
previous Phase 1 version. Fuel maps are one of the most important
elements in simulation-based models and the new feature of using
actual maps and drivetrain parameters would make the results more
realistic, allow the model to capture the effects of matching engine
and driveline, and ideally promote right sizing of the engines to
application. Different transmission options are included in the
model. The shifting behavior is now more realistic since is based
on both throttle and speed inputs, and includes the effects of a
clutch friction model. Phase 1 GEM shifting strategy was based
only on vehicle speed and there was no torque interruption during
shifting. Road grade has a major impact on fuel consumption and
its addition to the tests cycles would also make the results more
realistic. The treatment of vocational technologies, which were
limited to the tires in Phase 1, is considerably enhanced. The
approach followed by EPA to tackle the diversity of vocational
truck applications is appropriate. A few drive cycles are simulated
(ARE transient, 55 mph and 65 mph cruise with grade, and a new
idle cycle) and weighted differently based on specific application.
EPA's approach involved a good amount of testing and validation.
It must be said that most of these validation efforts only covered
fuel consumption results from reduced weight, better
aerodynamics, and better tires. The validation effort for
transmission types and engine- transmission-vehicle interaction
business information (CBI}. The agencies
cannot release information claimed as CBI
by manufacturers to the public.
21
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
was less comprehensive. However, based on the results presented,
Phase 2 GEM model would be accurate enough to support
regulation and drive technology adoption.
More broadly, I make one final comment on how the overall
modeling approach may meet EPA's overall goals for the
regulation, related to the public release of the GEM input and
output data. The existing and Phase 2 heavy-duty vehicle
regulation approach relies on the GEM inputs and outputs to
determine compliance. The GEM data are analogous to the light-
duty vehicle gram/mile1, heavy-duty vehicle gram/brake-
horsepower2, and light-duty vehicle mile-per- gallon3 compliance
values. For the heavy-duty use of the GEM in the greenhouse gas
emission regulatory program to meet the agency's own standard,
the input and output data from GEM would ideally be made
publicly available just as the regulatory data for the other
regulations for each engine and vehicle. For the public to have
confidence in the regulatory program that is built on a mix of
engine- and vehicle-model-specific inputs and modeled GEM
outputs, the underlying data would be presented in full in
downloadable data files (e.g., in Excel) as in other EPA
regulations.
References:
1 EPA http://www.epa.gov/otaq/crttst.htm
2 EPA http://www.epa.gov/otaq/certdata.htm
EPA http://www.epa.gov/otaq/tcldata.htm,
22
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer 3
Reviewer 4
Reviewer Comment
http://www.fueleconomy.gov/feg/download.shtml
The model appears to meet the stated purpose for which it was
intended, which is the prediction of integrated cycle-based vehicle
fuel efficiency and carbon dioxide emissions for vehicles with
preselected physical and drivetrain attributes. While this was the
subject of some detailed explanation in the model documentation,
the assumption of quasi-steady engine fueling and its extension to
fully transient engine operation is not without complexity in the
assessment of its validity. A further, acknowledged inadequacy of
the model in its current form is the limited ability of the user to
modify specific vehicle attributes, and component values and
efficiencies.
EPA's overall approach to meet the agency's compliance
requirements consists of making a validated simulation model
(GEM -II) available to OEMs so that they can check the
compliance of their vehicles against the agency's guidelines. The
extensive validation of GEM - II against dynamometer testing of
the actual vehicle shows a good correlation between simulation and
EPA Response to Comment
The agencies agree with the comments, and we
fully recognize this limitation of the model in
terms of the ability of the user to modify some
vehicle attributes. As explained in the
supporting document "Vehicle Simulation
Model" and Draft RIA Chapter 4 the
certification is conducted on a relative basis,
and the report shows relative errors in the range
of 2-3%, which is not perfect, but it is adequate
for the purpose of certification.
User inputs are limited to reduce complexity
and to provide a more common basis for
determining regulatory stringencies and to
reduce the expense and number of test
procedures required to provide accurate inputs
to the model. The agencies are seeking
comment in the preamble of the HD Phase 2
NPRM regarding the inputs to GEM and the
test procedures associated with them.
GEM PID controllers are not tunable modules,
meaning that the user has no option to change
any constants associated with this model. At
this time, we have validated GEM against 130
vehicle variants without noticing any issues
23
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
hardware to within ±5%. Most validations are well within ±3%.
This provides the user with a high level of confidence that the
physics has been correctly implemented and there are no
unresolved "bugs" in GEM-II.
Further, an additional level of confidence is achieved, with select
validation from four representative vehicle classes, namely: Class 8
Kenworth T700 truck, the Class 6 Ford F650 tow truck, the Class 6
box truck, and the New Flyer Refuse truck. There is a very good
correlation between GEM-II predicted engine speeds and
transmission gear shifting versus the same on the actual vehicle.
Recommendation: It is recommended that representative vehicles
from each of the classes be modeled in GEM-II and validation
results presented similar to 1.3.1 of the GEM-II Manual. It is the
understanding of the reviewer, based on the provided manual that
the following trucks were tested on a vehicle chassis: Class 6
Kenworth T270, Class 6 Ford F650, Class 8 Kenworth T700, Class
8 Cascadia Line Haul truck, and Class 8 New Flyer refuse truck.
Issues/Results that were not addressed by GEM-II:
During this review, the following unaddressed issues were
identified for GEM-II:
• GEM-II includes several PID controllers within its overall
structure. For example, the engine idle speed controller of
associated with PID controller. It should also
be noted that many of the PID controllers, like
the engine idle speed control, are scaled by
non-user-defined parameters like engine inertia
or vehicle mass to match their application.
The planned interface which manufacturers will
use to certify via GEM does not provide the
option to output plots. The functionality is not
considered necessary to complete the
compliance requirements identified in the rule.
Thermal modeling is always an issue. Ideally,
it should be included in some form. However,
it is challenging to develop a correlation to
qualitatively predict the thermal impact
associated with cooling and transmission heat
rejection in terms of a certification tool. This
development can be expensive and time
consuming. This could place an additional
burden on manufacturers who must submit
certifiable reports to the agencies, which
complicates the process of the certification.
Mechanical loads can be also modeled through
a simple correlation. However, similar to the
thermal modeling issue, this could place an
24
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
the engine is implemented as a PID with three gains. The
unaddressed issue in this regard is stability feedback to the
user caused by unrealistic hunting for the idle speed. What
safe-guards are in place within GEM to inform the user of
clutch chattering since GEM does not model second order
inertial effects caused during clutch engagement.
It is possible for the vehicle not to meet the driving cycle as
a result of excessive grade or weight or other issues with
the transmission/engine. The unaddressed issue in this
regard is a feedback alert to the user during the time
instances when the vehicle is significantly slowed down
and does not meet the desired driving cycle. The output file
does not alert the user on the number of time instances
when vehicle tracking was compromised.
GEM-II does not address thermal characteristics of the
engine cooling system or the heat rejection of the
transmission fluids. These thermal issues affect the
operational duty-cycle of the engine fan, which will affect
the fuel economy. At present, GEM-II models the parasitic
loads as a constant average number.
Recommendation: Allow the user the ability to introduce an
engine load dependent mechanical accessory curve which is
more realistic than a constant average number. A simple
heat model may be used to capture the effect of thermal
characteristics of the multiple radiators in a typical MD and
HD engine.
Although GEM-II does not model tire slip/lockup during a
additional burden on manufacturers who must
submit certifiable reports to the agencies, which
complicates the process of the certification
The supporting document, "Vehicle Simulation
Model", does show some detailed comparisons
in a sub-model level, such as engine and
transmission. It is extremely time consuming
to show all 130 vehicle variants at such a
detailed level. Furthermore, certification only
requires composite and weighted values, and
therefore, the program should serve this
requirement as long as those composite and
weighted values are correlated well against the
testing data.
In terms of transmission shifting, all
independent transmission manufacturers we
contacted are reluctant to share their shifting
strategies with the vehicle manufacturers due to
concerns over proprietary information.
Therefore, the agencies developed their own
generic shifting algorithm for modeling
purposes. If GEM under-predicts the benefits
of the transmission, manufacturers have the
option to conduct powertrain testing where an
engine and transmission are evaluated in a
25
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 2. Charge Question 1
Charge Question 1: Please comment on EPA's overall approach to the stated purpose of the model (meet agencies' compliance
requirements) and whether the particular attributes found in the resulting model embodies that purpose. Were there critical results or
issues that were not discussed or addressed by the GEM tool or its component sections?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
hard deceleration, the effect of ignoring this on fuel
economy is negligible.
The validation results included kinematic comparisons
(speeds, gear number) between GEM-II and the actual
vehicle on a chassis dynamometer. While, the kinematic
comparisons look very favorable, the dynamic comparisons
(engine load and engine fueling) are missing in the results
section of GEM-II.
The transmission shift strategy can affect fuel economy and
emissions. GEM-II allows the user to preselect different
transmission types (manual, automatic or automated-
manual). However, it was not clear how to modify the
GEM-II default shift strategy with an OEM proprietary
shift strategy.
Although GEM-II results have been extensively validated
against dynamometer test data in a controlled lab
environment, it is unclear how well GEM-II will compare
against real world road testing, especially with temperature
fluctuations. For example, the lack of a thermal model in
the engine model may cause GEM-II results to deviate from
on-the-road test data, where the engine fan is cycled on and
off based on thermal loads on the engine. Each time, the
engine fan turns on, fuel economy is affected.
powertrain dyno cell.
The agencies fully recognize the importance of
the GEM validation against real-world vehicle
operations. However, it is expensive and time
consuming to launch such a program.
Currently, the agencies are working with SwRI
to conduct this task in the hope that full GEM
validation against real-world vehicle operations
can be released to public in the second half of
2015.
26
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
a) Elements in each system used to describe different vehicle categories;
Reviewer
Name
Reviewer Comment
EPA Response to Comment
Reviewer 1
The proposed GEM Phase II model represents a substantial
advance over the model used to implement the first phase of
truck efficiency legislation, and encourages more technology
advances from manufacturers in consequence. Improvements in
truck efficiency are based primarily on reductions in
aerodynamic drag, tire rolling resistance and engine brake
specific fuel consumption, and this was recognized in the first
GEM model. Practically, there is less to be gained from
aerodynamic improvements in most vocational truck operation
than in long haul trucking and the GEM model as presented
neglects vocational truck aerodynamics, and the modelers are
right to exclude aerodynamic parameter entries for low speed
trucks. However, the overall GEM structure is capable of
modeling aerodynamic improvements for niche vocational
designs, and has the flexibility to extend beyond the present
exclusion of the drag coefficient. In this way, the capabilities of
the model, as received, will be far greater than the executable
version that is finally used for compliance.
A major theme in the industry is that efficiency gains are
significant from design integration, particularly powertrain
integration. But it is understood that combined powertrain
control is proprietary. The supporting language might address
this more clearly, noting that the GEM model employs just
steady-state maps and a defined set of gear ratios, and cannot
predict the benefits of more sophisticated integration. In a
The comments on gasoline and natural gas engines
are well taken. The agencies are currently
conducting a program at SwRI to address gasoline
engine performance related to the rulemaking. We
are also actively collecting the engine performance
data on both types of engines from manufacturers.
The GEM Phase 2 shift algorithm is based on the
torque curve and fuel map of the engine and such
will adapt shift points uniquely for every engine
map provided, regardless of fuel type.
Manufacturers will have the option to perform
powertrain testing to account for improvements,
such as those mentioned by the reviewer.
The agencies are seeking comment on whether to
include GEM inputs for vocational vehicle
aerodynamics in the preamble to the HD Phase 2
NPRM.
27
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
similar fashion GEM cannot predict the benefits of learning
algorithms, look-ahead strategies and intelligent vehicle
systems for the optimization of powertrain efficiency on
specific routes. These are emerging approaches, but it is
acknowledged that it would take great effort to configure GEM
to deal with these details and it would be difficult to assure
their generic benefit in revenue service. GEM has some check-
a-box options proposed for features that cannot be modeled.
The model, as provided, was oriented to diesel engines. The
shift strategies also considered the engine torque curve for
execution. Naturally aspirated gasoline, boosted gasoline and
natural gas engines are likely players in the next five years and
may warrant separate and careful consideration because their
characteristics, torque curves and efficiency maps differ
substantially from the diesel engine properties.
Reviewer 2
Overall the model structure and its systems are appropriate and,
in large part, complete. Generally, the performance of each
component model and the underlying equations and physical
principles are valid throughout (see some finer details below).
The input and output structures interact with the model to
obtain the expected result in a way that is sound. The following
sub-sections comment on specific issues regarding model
structure, individual systems, as well as default values, in no
particular order of importance.
Fixed pavloads
Phase 1 GEM had predefined engines, driveline parameters,
and payloads for every category. An issue that may arise when
Fixed Payload
The agencies are fully aware of the technical issues
related to payload. However, allowing payload as
a variable means that agencies must develop a
standard that varies with payload, which would
complicate the rulemaking, specifically the process
of certification. Furthermore, it would be
challenging to verify the in-use payload of a
vehicle application, making audit challenging. In
addition, this approach would force tractors
running without trailer, or bobtail, to be considered
28
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
using user-defined engine fueling maps in combination with
predefined payloads is that some simulated vehicles, with lower
power-to-weight ratios, will show higher deviations from the
target speed-distance trace. This affects the simulation results
since these underpowered vehicles will take more time to
complete the assigned route and will show a lower average
speed. This could lead to underpowered vehicles being
improperly credited.
Appropriate matching of engine, transmission gear ratios, axle
ratios, and tire radius is only going to be promoted if the GEM
payloads closely match actual vehicle operation. Right sizing of
powertrains to application does not seem to be promoted when
payloads are predefined for a particular vehicle category. In
order to recognize engine power matching to vehicle road load,
payload needs to be a user input rather than a predefined
parameter. The regulatory approach and modeling would
ideally recognize and promote market diversity and identify
potential discrepancies between actual payloads and GEM
payloads. There is an existing trend towards smaller engines,
but also some applications require larger engines. On the other
hand, if the truck manufacturer is allowed to input vehicle-
specific payloads, some issues may arise in terms of
enforceability (How do the regulatory agencies ensure that the
vehicles are operated close to the payload values at which they
were certified?), that may also open the door for the
manufacturers to report numbers for their own benefit, and
adds complexity.
out of compliance because they are not carrying the
declared payload. Therefore, the agencies simplify
the process by proposing fixed payload values.
Pull-Down Technologies
Pull-down technologies are also known as the
technology improvement inputs for the rulemaking.
The supporting document, "Vehicle Simulation
Model" and the HD Phase 2 NPRM preamble
describe many aspects of the pull-down
technologies. We also seek comment on whether
the technology improvement inputs should be in
terms of percent reduction or absolute grams of
COz per ton-mile. Basically, the agencies have
been actively talking to all relevant manufacturers
regarding these technologies as selectable items.
We proposed a conservative approach recognizing
these potential technologies. All the technologies
considered as pull-downs would be those that GEM
would not be able to model or are not fully
recognized over the limited certification drive
cycles. Technologies mentioned by this reviewer,
such as electric coolant pumps would only show a
partial benefit in the engine fuel mapping process,
and therefore the pull-down item associated with it
accounts for the other remainder of the benefit seen
on the road.
29
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
An option could be to adjust the payload on a few pre-defined
bins based on certain parameters that are indicative of vehicle
road load (e.g. engine displacement, engine power, final drive
ratio). Under this option, a performance criterion that captures
the trace-following capabilities of the simulated truck (e.g. a set
threshold of percent difference between target speed and
simulated speed) can be used to force certain engine-vehicle
combinations to switch to a lower payload bin if they don't
follow the trace according to the specified criterion. Another
option would be to impose a CO2 penalty based on the ratio of
simulated average speed to target average speed. Ideally, the
allowed deviations from the target trace should be minimized
for the simulations to be considered valid and allow
comparisons between them.
Drop-down technologies
The agencies have identified a list of technologies that provide
fuel consumption benefits but are difficult to simulate
accurately. They are developing feature-based drop-down
menus that make post-simulation adjustments (percent
reductions) to the results. It appears that manufacturers have
not taken much advantage of the Phase 1 advanced technology
structure to earn credits so it is important to try to include most
of the technologies in some way. However, drop-down menus
inherently assume that all the technology variants within a
technology category provide the same fuel consumption
benefits. Not all the models and brands of a certain technology
feature would provide the same fuel consumption benefits.
There is the risk of giving artificial credits to products that
Although these values used by GEM are fixed as
default values, the user does have the option to use
off-cycle credit proposed by the rule, similar to the
innovative credits in Phase 1, to quantify the
additional benefits of individual technology.
Driver Subsystem
All the driver model related constants are not
tunable. We tested these pre-selected PI controller
related constants against over 130 vehicle variants
without any noticeable issues. In addition, the
driver controller constants are scaled by vehicle
mass and therefore adjust automatically for each
simulation run.
Transmission
The auto-shifting tables for all three types of
transmissions are different in the form of internal
constants which are not user-tunable. The
supporting document, "Vehicle Simulation
Model", and HD Phase 2 NPRM Draft RIA
Chapter 4 includes a table to show the impacts of
the shift algorithm on overall vehicle performance
as opposed to using manufacturer-supplied shift
tables. We agree with the reviewer that we need to
provide a clearer description of this subject and it
will be clarified in Draft RIA Chapter 4.
30
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
perform at a lower level than the value that is selected from the
drop- down menu, thus rewarding poor performers. Also,
technology products with better than average levels of
performance would not get additional credits, which is a
disincentive to make investments in the development of such
technologies. The default improvement values (percent
reductions) developed by the agencies were not shared for this
peer review but they are of relevance and need to be
determined with care. Currently, the users have no flexibility to
enter their own values. Giving the users the flexibility to enter
their own values (after testing and with proper documentation)
could offer a way to reward good performers.
It seems that applying adjustment factors in terms of percent
reductions rather than applying predefined credits in units of
go2/ton-mile or gal/ton-mile may punish good performers.
Assuming that truck A emits 90 go2/ton-mile and truck B emits
100 gCO2/ton-mile. If a certain technology improvement value
is set at 5%, and both trucks use such technology, truck A
would get 4.5 gCO2/ton-mile credit and truck B would get 5
gCO2/ton-mile credit. This discrepancy of incentives can
exacerbate if the trucks use more than one drop-down
technology and the agencies decide that the percent
improvements are additive. So it would be good for the
agencies to support whether and why percentage-based (versus
gCO2/ton-mile based) are most appropriate. Also the agencies
might address, in such drop-down menus whether such
technology improvements are indeed additive or not.
Another issue with drop-down technologies is that there is the
We fully agree with the reviewer's comments on
the powertrain test. As a matter of fact, the
powertrain test is one of the options that
manufacturers can use to address benefits GEM is
unable to fully capture.
Engine Fuel maps
The proposed engine fuel mapping procedure is
detailed in the proposed regulations in 40 CFR part
1036.
It is always challenging to use a steady state map
approach to account for transient operation. While
there are many ways that the vehicle model can be
improved for those behaviors, there are always
trade-offs in terms of computational speed and
accuracy. Furthermore, including more advanced
models, such as model based control could
substantially improve accuracy, but the collection
of test data plus calibration of the model against
the data would be beyond the agencies'
capabilities, and could be expensive, time
consuming and error-prone. This kind of advanced
modeling could take much longer to run as
opposed to the proposed executable version of
GEM, which only takes a few seconds to complete
one certification vehicle. It is very typical for a
vehicle manufacturer to run thousands of
31
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
potential for double counting of technology benefits. As an
example, an electric coolant pump is listed as a drop-down
technology. Depending on the engine mapping process, the
resultant engine fuel map may already capture the benefits from
that technology. Running a simulation with such a map, and
later improving the results using a drop-down menu will double
count the benefits. If EPA could respond to how potential
double-counting situations are minimal, that would be helpful.
Driver subsystem
In vehicle simulation modeling, it would seem that the driver
ideally would be excluded entirely as a factor that could
influence the GEM regulatory compliance results. Using the
same driver model for all the vehicles seems to be an
appropriate choice. However, additional documentation is
needed for this subsystem. There are no details about how the
proportional and integral gains of the PI controller have been
selected. Are they representative of current drivers? Are they
tuned to enhance the trace-following capabilities of the model?
The look- ahead feature also lacks documentation. Is it bringing
any advantage to the trace-following capabilities of the model?
How was the time span value for such feature selected? Ideally
EPA would provide some consideration and discussion of such
factors to provide greater assurance that no anomalies occur in
compliance results from company-to-company technology
strategies as well as tested-versus-real-world results for the
relative technology benefits.
Transmission subsystem
simulations for certification. It is not practical to
introduce such complicated modeling processes to
perform certification at this time.
We propose a single transient factor in the HD
Phase 2 NPRM, as the reviewer recommended.
We are also seeking comment on the transient
correction factor.
Modeling of idle cycle
The HD Phase 2 NPRM described in Chapter 2 of
the Draft RIA and Preamble Section V provides
more detailed description on modeling of the
idle cycle.
Regarding "trace following" and the idle
calculation, for tractor-trailers the idle weight is
zero and the simulation grams/mile are
multiplied by target mph and also divided by
target mph so what remains is simulation
grams/mile which will reflect the modeled
performance (or under-performance] of the
vehicle.
For vocational vehicles the same is true with
regards to the simulation grams/mile over the drive
cycles. Idle consumption takes places at zero
speed and is measured in grams/hour so there is a
32
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
There are some transmission-related features that are confusing
and need to be clarified. The report mentions that the different
transmission models: manual (MT), automated manual (AMT),
and torque converter automatic (AT) are built of similar
components, but each features a unique control algorithm.
However, the model seems to use the same "auto shift
algorithm" to determine the operating gear for any transmission
type. The differences in the control algorithm of the three
different transmissions are not clear and need to be provided.
Since transmissions are an important new addition for Phase 2
GEM, it is important to let the reader know that the control
strategy (e.g. shift points) or the selection of predefined
transmission parameters (e.g. efficiencies and inertias at
different gears) are not creating any artificial advantage of one
technology type over the others. I suggest presenting a
comparison of the same simulated truck with different
transmission types. It is also important to highlight in the report
that the new transmission controller is based on both speed and
throttle position, and differs from the Phase
1 transmission controller, which was solely based on vehicle
speed. The rule-based approach of the "auto shift algorithm"
would ideally be documented.
It would be appropriate for the agencies to acknowledge that
Phase 2 GEM simulations can capture some but not all of the
benefits of powertrain integration. The simulation would
adequately capture engine down speeding since the users have
to input specific transmission gear ratios, final drive ratio, and
tire radius. However, there are many complexities in the control
conversion factor required to obtain grams/mile.
The target weighted average speed represents that
conversion factor and does not alter the modeled
vehicle performance (or under-performance) over
the drive cycles.
Trailer
The HD Phase 2 NPRM described in Chapter 2 of
the Draft RIA and Preamble Section VI provides
more detailed description on how trailers are
handled. Tractor manufacturers determine the
coefficient of drag area for a tractor-trailer
combination. The trailer used in this
determination is a "reference trailer" that is
specified in the regulations (40 CFR part 1037}.
Details of the test procedures for the tractors
are included in the HD Phase 2 NPRM Section III.
Accessories
The agencies chose the same approach as Phase 1
to model accessories, mainly because it is not an
easy task to model accessory improvements, which
requires time consuming and expensive testing and
validation of the model. Allowing user input of
accessory loads would require each user to know
ahead of time the expected load for each vehicle in
use and while potentially providing more accurate
33
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
strategy when it comes to integrating engine and transmission.
Integrated engine-transmission powertrain approaches with
advanced controls and shifting algorithms that many companies
are developing could result in significantly more (or less)
benefit than the agencies determine as the appropriate default
emission-reduction effect.
As an example, if two different vehicles have the same
driveline parameters (tire radius, final drive ratio, transmission
gear ratios, transmission inertias, and transmission efficiencies)
and AMT transmissions from different manufacturers, they will
obtain the same simulation results in GEM but, due to differing
control strategies and other design characteristics, they will
show different fuel consumption benefits in reality. It cannot be
expected that all the AMT transmissions bring the same fuel
consumption benefits. The drop-down menu option won't
handle these differences unless there is an option to choose
manufacturer-specific transmissions or otherwise input such
data.
As a result, there is an opportunity here to leverage the
powertrain testing and provide the option for manufacturers to
better capture the fuel efficiency gains coming from the control
strategies and other complexities that are not adequately
captured in GEM. Another advantage of powertrain testing is
that the manufacturers would not need to disclose confidential
information. The results from powertrain testing can then be
implemented as correction factors for the GEM results. Using
correction factors, GEM results could be multiplied by a fixed
results would place an unreasonable burden on the
user and manufacturers. If the accessory is
normally part of the engine, the engine mapping
conducted on the dynamometer should be able to
account for some of those losses, thus being
modeled through the engine fuel map. If accessory
improvement comes from the vehicle, and GEM is
unable to model it, manufacturers can either use
pull-down technology or use innovative credit to
recognize these accessories.
34
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
percent improvement obtained by comparing the results of
powertrain test and GEM simulations under the same torque-
speed trace.4 The default benefits for transmission
improvements would ideally be set to be appropriately
conservative (i.e., lowest expected value based on various
industry results) in GEM. The drop-down menu could still then
be offered as a default, for the manufacturers that decide not to
use the powertrain testing. Then, for the powertrain option,
companies would ideally be provided clear testing procedures
and guidance to demonstrate the emission-reduction impact of
their advanced powertrain approaches with physical vehicle
testing in simulated real-world conditions.
Engine fueling maps
The inclusion of manufacturer-specific engine maps is a critical
feature to reflect company differences and detailed engine-
specific characteristics that reflect real-world fuel consumption
and emissions. This is an important addition to GEM, but there
is lack of documentation of the engine mapping procedure. I
imagine that a fairly prescriptive procedure (including number
of points, preconditioning and warming procedures, fuel
properties, etc.) is described somewhere else in the larger
regulatory development document but this chapter would
ideally include a brief description of the procedure so the
reader knows which engine accessories are included or
excluded during the engine mapping procedure.
It is noted that there are many advanced features that may
affect fueling but are not captured by using a steady-state fuel
35
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
map. Manufacturers are going away from traditional map-based
strategies and are going towards model-based controls. Diverse
thermal management strategies are utilized, and some engines
use dual torque curves. Have the agencies considered how to
handle these technologies? This could have important
implications for how tested steady-state engine maps, and GEM
modeling, and real-world emissions characteristics could differ.
As a result, we recommend that the agencies discuss such
industry approaches in the rulemaking and investigate ways to
ensure that tested results are aligned with real-world engine and
vehicle operation the results in fuel consumption and
emissions.
The approach used to quantify the transient correction factor
(run GEM with the engine map, then use the torque-speed
points in the engine dynamometer and compare measured
versus simulated results) is appropriate. Ideally the transient
correction factor may be obtained for each individual engine.
However, since there is a need for selection of a vehicle in
GEM in order to get the torque-speed trace. It would become a
hard task for the agencies to try and run a transient correction
factor for each vehicle-engine configuration. For practicality, I
recommend provisionally using a single correction factor and
maintaining the option to refine it over the years with
additional testing.
Modeling of idle cycle
The idle cycle modeling would gain from increased
documentation. Using a gCO2/mile value for an idle cycle at
36
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
first seems counterintuitive (i.e., there are no miles traveled) so
a complete description of the calculation method would clarify.
It would be desirable to present some validation results for the
idle cycle modeled in GEM compared to experimental results.
Some of the engine auxiliaries may not be enabled while doing
the test, and the map could be underestimating actual idle speed
fueling rates. There are also engine thermal management
strategies that are used to keep appropriate after treatment
system temperatures. These strategies vary from manufacturer
to manufacturer and could increase idle fueling substantially.
The "trace following" issue discussed above also has
implications in the calculation of idle cycle g/mile value. For
this calculation the fuel rate in units of grams per hour [g/h] is
converted to units of grams per mile [g/mile] using the
weighted average speed over the three non-idle cycles. The
target speed is used for this calculation and not the actual
simulated speed, which may penalize smaller engines. I suggest
EPA to consider if this issue might be significant.
Trailers
Although there is a parameter in GEM for trailer tires' rolling
resistance, it is not clear how trailer aerodynamics is going to
be modeled in GEM. Trailer aerodynamics can bring about
two-thirds of tractor-trailer aerodynamic benefits, so this is a
critical area that requires documentation and specification of
the procedures for the vetting, binning, and including the input
data. My understanding is that the Coda input parameter is for
the tractor only (mid-roof and low roof tractors are tested in its
37
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
bobtail configuration), or for the tractor using a "reference" 53-
ft dry van trailer (for high-roof tractors coast-down test).
Trailer aerodynamic devices can reduce the overall tractor-
trailer combination aerodynamic drag and ideally the Coda
used in simulation should represent the combination. It seems
that there is no current provision to include the effect of trailer
aerodynamics as an input in GEM. The report needs to clarify
how the GEM model is handling trailer parameters (including
aerodynamics, tires rolling resistance, and weight reduction)
and if the model is going to use a predefined "reference" trailer
for all the tractors. Ideally agencies would give credit to tractor-
trailer integrated designs although it would be difficult for the
agencies to ensure in-use compliance of matching of tractors
and trailers.
Accessories
There are opportunities for fuel savings from mechanical
accessories and electric accessories but the agencies decided to
keep with the Phase 1 approach of having pre-defined and not
customizable power from accessories. If these parameters are
assigned default values, there are no incentives to implement
new technologies that could have greater impact. Allowing
accessories power consumption to be user-defined inputs can
be used to promote developments in technologies that reduce
the power requirements of accessories such as the alternator,
air-conditioning compressor, power steering pump, or cooling
fan. There are other opportunities for engine accessories such
as oil, coolant, and fuel pumps, but is not clear at this point if
all those savings are going to be captured by the engine
38
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
mapping process.
References:
4 See Sharpe, Delgado, Muncrief (2015) Comparative
assessment of heavy-duty vehicle regulatory design options
for U.S. greenhouse gas and efficiency regulation.
http://www.theicct.org/us-phase2-hdv-regulation-design-
options
Reviewer:
The elements in each of the systems (engine, transmission,
axle, vehicle attributes etc.) seem appropriate and complete.
The specific selection of the engines and transmissions chosen
will cover a large portion of the current heavy-duty vehicle
fleet, although of course any specific, single selection of
powertrain hardware or powertrain hardware attributes
necessarily limits the range of vehicles that can be simulated
with that same selection.
Non-conventional or alternative powertrains can be
certified through powertrain testing.
Reviewer 4
The three main powertrain components that can affect fuel
economy and greenhouse gas emissions in a vehicle are: engine
fuel map, transmission type and efficiency map, and vehicle
aerodynamic improvements (including tire rolling friction
improvements and weight reduction technologies). In this
regard, GEM-II addresses all the aforementioned components
by providing steady state maps for each powertrain component,
which an informed user can change to represent specific
technology improvements. GEM-II comes with certain standard
transmission models, namely: manual, automatic, and
automated manual transmissions. The user would select the
appropriate transmission and GEM-II would automatically
select the user-specified transmission.
The proposed Phase 2 version of GEM for
certification is an executable version code, which
does not require Matlab/Simulink license. In
additional, a plain text formatted file will be used
for user inputs. There is no need for user to
understand the coding structure of GEM in a
Matlab/Simulink format. Neither manufacturers
nor users will be able to modify the structure of
GEM in any way. The Phase 2 GEM User Guide,
provided with the NPRM, will provide details on
how to use GEM.
A technology improvement such as a proprietary
39
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
Recommendation: It is recommended that additional
instructions are provided if the user wants to change the engine
map. Ideally, this would be done from a user specified
spreadsheet in a GEM-II compatible format, since the user may
not be fluent in Matlab. In this regard, a clear explanation of all
the variables used in GEM-II would also help significantly. For
example, it was not clear how to change the transmission shift
schedule if an OEM chose to do so. Further, since GEM-II is
modular with hierarchical layout of component layers, it is
challenging for an OEM user to insert a technology
improvement deep within one layer and not affect the layers
above or the execution of GEM-II. It is not clear, from this
initial review, how a technology improvement such as a
proprietary transmission shift schedule can be evaluated in
terms of gains in fuel economy and greenhouse gases. This
comment applies to other technology improvements as well,
such as partial engine cylinder deactivation (power on demand)
or electrification of certain mechanical accessories.
The vehicle categories that GEM-II addresses range from Class
2B to Class 8 HD conventional vehicle powertrains. This is
achieved by four root-level systems in GEM-II at the root level,
namely: the ambient, driver, vehicle, and powertrain modules.
Each of the aforementioned modules consists of several sub-
modules organized in a hierarchical manner. Each root-level
module outputs a data bus that is mixed into a single data bus.
The aforementioned four main systems of GEM-II correspond
to the four main components of a HP vehicle, namely: driver,
transmission shift schedule would not be
considered under the current GEM. Rather, the
agencies encourage the user to use the powertrain
option described in Chapter 3 of the HD Phase 2
NPRM RIA to address any benefits associated with
the transmission and its integration with the engine
that GEM is unable to model.
40
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
ambient conditions, vehicle chassis and powertrain modules.
This one-to-one correspondence between the root-level GEM-II
models and an actual HD vehicle makes for an easily
understandable structure. Further, the modular organization of
the GEM-II contributes to easier debugging and isolation of a
numerical problem during simulation.
The powertrain module is the most populated module in GEM-
II. It contains the engine, transmission and driveline sub-
modules and accessories (mechanical and electrical). The flow
of data information corresponds to an actual vehicle
powertrain, with the engine output driving the transmission,
which in turn drives the driveline components.
The modular layout of GEM-II, its correspondence with a real
conventional vehicle is therefore appropriate and complete for
the reasons stated above. The modular structure and the
hierarchical arrangement of modules to mimic a real vehicle
system makes the integration of additional modules and
capabilities easier to implement. The signals are clearly marked
and follow a logical naming convention that facilitates the
addition of additional modules and capabilities into GEM-II.
b) Performance of each component model including the reviewer's assessment of the underlying equations and/or physical principles
coded into that component;
Reviewer 1
MATLAB/Simulink remains an excellent basis for the GEM
model. It is well suited for the exploratory and development
framework, as well as the production of a more limited
executable model.
The agencies use their own auto-shift algorithm for
all transmissions for many reasons as described by
the supporting document - "Vehicle Simulation
Model". We recognize that this is not an ideal
modeling approach, but we do offer the powertrain
41
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
GEM may be viewed on three levels. At the highest level, the
MATLAB/Simulink platform allows GEM to be anything it
chooses, with the addition or alteration of modules and
component models. At the second level, there is the model
provided for review, which has the innate ability to deal with a
wide range of cycles and truck configurations. At the lowest
level will be the final executable version, where certain
parameters are fixed, and where the duty cycles are chosen for
determination of compliance. Only the second and lowest
levels can be considered in this review.
This GEM model is being produced at a time where engine
control strategy development and integrated powertrain
controls are advancing rapidly. Also, transmission options are
now far wider than in the earlier GEM model, where
unsynchronized manual, synchronized manual and traditional
automatic transmissions dominated the marketplace. GEM is
challenged in modeling and giving credit for the technology
subtleties that will emerge in the marketplace over the next five
years.
The use of distance, rather than time, as a basis for the test
cycles represents a great and important advance. It was widely
recognized two decades ago that light duty automobiles were
capable of far more aggressive performance than was embodied
in the FTP-75, although the FTP-75 was used as the norm, and
still embodied allowances for vehicles that could not follow its
modest accelerations. In the heavy-duty arena sustained use of
full engine power both on grades and during acceleration is the
norm. All else being equal, a more powerful truck will
testing approach that allows manufacturers to
quantify the benefits associated with their own
more advanced shifting technologies and
integration benefits with the engine.
We are fully aware of the limitations of the steady
state engine fuel maps that are proposed for GEM.
We proposed the concept of the transient correction
factor for the transient cycle to minimize the
deficiency of the steady state map. We are also
aware that this transient correction factor is derived
from the diesel engine test, where gasoline and
natural gas engines may behave differently. We
are currently asking for comments on this
approach. We will make refinements in our final
rulemaking, as needed based on the comments we
receive. In addition, manufacturers have the option
to conduct powertrain testing to demonstrate their
engine's transient performance.
At this time, we are proposing a single value for
axle mechanical efficiency, and ask for comments
of whether we should use a look-up table to model
axle efficiency based on axle efficiency test results.
The agencies will make appropriate adjustments in
the final rule based on the comments received.
It is not straight forward to model tire rolling
resistance as function of speed without
42
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
complete its duties in less time than an underpowered vehicle,
and often the more powerful choice represents the overall
economic optimum. However, the more powerful truck will
spend less time at full power, and will enjoy a reduced average
"%load" in revenue service. The adoption of the distance-based
approach is an important step toward matching the GEM model
to real-would use. However, as standards are finalized, the
accelerations and grades of cycles embodied in the GEM
execution must represent real life as well. If the grades and
acceleration values are not appropriately challenging, the
distance-based approach will look more like a time-based
approach, because underpowered trucks will be able to follow
the trace in the minimum time allowed. This would divorce the
market incentives from the environmental incentives and create
a false impression of fuel efficiency capability at the expense of
economic reality. It is important that the grades and test
weights used are realistic and representative. Further, the
indication of the ratio (actual / minimum time for cycle
completion) is a very beneficial output value.
A driver's habits are known to have a measurable impact on
truck fuel efficiency. This is in part due to gear selection that
influences the engine operating envelope, and in part due to
transient pedal demand (driver PI controller), that may cause an
engine to depart from steady-state mode to a greater or lesser
degree. The GEM load demand controller and GEM gear
selection algorithms can be configured to reflect different
driving habits. If the chosen GEM driver yields a better fuel
efficiency than would be expected of the national average
comprehensive tire data. Besides, just like the
comments made by this reviewer, detailed tire data
of this kind are not in the public domain and test
methods are not universally defined. It is
challenging for the agencies to develop a more
advanced tire model at this time.
The agencies present details of the proposed road
grade cycles in the HD Phase 2 NPRM preamble
and welcome comment on the representativeness of
these and other road grade profiles developed. The
agencies will reflect the final road grade profiles in
the final version of the HD Phase 2 GEM.
43
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
driver, then GEM will fail to grant auto-shifting technology and
certain engine control strategies their full potential contribution
to efficiency gain. It is recommended that this issue is at least
explored to provide driver sensitivity results to the
accompanying document.
The engine model in the new GEM is still a steady-state map.
The documentation acknowledges that transient fuel efficiency
will differ from steady-state efficiency at the same speed and
load in time. A transient adjustment factor (TAP) seems likely
for the final version, but the use of a steady-state map with a
TAP does not encourage manufacturers to improve transient
fuel efficiency if a generic TAP is assigned. This reviewer
appreciates that manufacturers are unwilling to reveal strategies
that represent a substantial fraction of product value, yet they
cannot take advantage of their improvements in the model
without this information. Two solutions are possible:
(i) Determine a manufacturer-specific TAP rather than an
agency-specified TAP, or
(ii) Determine an alternate fuel efficiency map for transient
(e.g. Heavy-Duty FTP) operation, and use that map or
both steady state and transient maps for the transient
cycle.
In addition, TAP will need to be considered carefully for
gasoline and natural gas engines, where enrichment
management has an important effect on fuel efficiency. In a
similar way, cylinder deactivation strategies may be difficult to
44
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
characterize for throttled engines, because the deactivation
strategy is not well represented by a steady-state map. It is
important to have a TAP strategy that acknowledges successes
of manufacturers in lowering transient operation disadvantages.
The EPA developers have sought comment on the issue of axle
efficiency. It can be dangerous to employ the overall fuel
economy computation to compare two approaches to modeling
a single component, particularly if that component represents a
small loss. In one case, the overall fuel economy differed by
1.67% when a simplified efficiency was compared with a
lookup table for rear axle efficiency. This 1.67% represented
more than a third of the loss associated with that component.
To place this in further perspective, the 1.67% can be compared
to the 3-5% contributions of a major component, such as side
skirts on a semi-trailer at freeway speeds. Generically, if a
component model with higher fidelity is available, in the
opinion of this reviewer, it is worth including the detail in
GEM, even if the component model is fixed in GEM.
Furthermore, the existence of such a model may provide a
clearer pathway for entering technology advances by a
manufacturer who improves that component in the future.
The tires in GEM are simply characterized by a rolling
resistance coefficient and an effective loaded radius. It is
known that the coefficient varies with vehicle speed and
contact area, while longitudinal slip can change the revolutions
per mile by a few percent under driving torque and braking,
which could bias results more substantially when grade is
45
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
present. This could be coded into GEM, but it is probably more
appropriate to treat it as a comment, since detailed tire data of
this kind are not in the public domain and test methods are not
universally defined.
Reviewer 2
See response for part A.
Reviewer:
One concern with the structure and form of the model is that a
steady-state engine fueling map is used in each case to simulate
transient engine operation and hence dynamic vehicle
operation. This is in addition to the simulation of operation
under nominally steady vehicle speeds or cruise operation. In
the case of nominally steady operation, the use of a steady-state
fueling map is well-justified, but the quasi-steady assumption
required to allow the extension of the use of such a map to
transient operation requires additional justification. Heavy duty
compression ignition engines have high rotational inertias (due
to the relatively high mass components required to survive the
high combustion pressures), high mechanical friction (due to
high effective compression ratios) and relatively slow air and
exhaust transfer processes (due to the excess air flow rates
accompanying their lean, un-throttled operation). In addition
they have relatively high thermal mass due to their large
physical mass required to withstand the stresses and strains
resulting from high combustion pressures. All of these features
conspire to result in deleterious combustion effects under
highly transient engine operation. In most cases the end effect
of these phenomena is to reduce the engine brake thermal
efficiency under transient modes, beyond that which would be
expected under steady or quasi-steady fueling operation. In
most cases the additional fuel that is required to undertake a
We are fully aware of the limitations of the steady
state engine fuel maps that are used for GEM. We
introduce the concept of the transient correction
factor for the transient cycle to minimize the
deficiency of the steady state map. We are also
aware that this transient correction factor is derived
from the diesel engine test, where gasoline and
natural gas engines may behave differently. We
are currently asking for comments on this
approach. We will make refinements in our final
rulemaking.
46
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
specific engine transient torque trajectory, beyond that
estimated using a quasi-steady fueling assumption, would
typically be less than 10% of the total integrated fueling, but in
most cases the effect on integrated fuel efficiency over transient
duty cycles is non-negligible.
In general, incorporating an additional component into the full
accounting of the vehicle load by over-accounting for the actual
total rotational inertia (in the form of an effective or "added"
mass or inertia) of the drivetrain and driveline system does
allow for the quasi-steady assumption to hold. However in
general a quasi-steady, forward-looking simulation such as is
used in this model, tends to under-predict the actual vehicle
energy usage under transient duty cycle operation. I notice that
this issue is addressed in Chapter 1.4.1.8 Transient Adjustment
Factor, but the designation of a single correction factor for a
specific engine or powertrain configuration is likely to be
unsuitable in some cases, and has the potential to cause
prediction inconsistencies. Note further that the required
correction factor might not be uniquely engine-specific, but
might vary for the same engine in different vehicle and
powertrain configurations.
Reviewer 4
GEM-II follows the model of a single wheel with a
concentrated mass at the center of the wheel. The physics
coded into the modules are based on Newton's second law of
motion for this concentrated mass. At the time of this review,
no other dynamic equations were found in GEM-II, except in
the clutch and torque converter. The engine, accessory,
transmission and driveline components are characterized by
The model physics are Newtonian, as noted, and
follow textbook equations for Newton's laws and
conservation of energy.
47
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
steady state maps. The equivalent mass of rotating inertia
components are also correctly included in the vehicle module
of GEM-II. Rotating inertias are correctly reflected
downstream to the tire. The inertia is converted to a virtual
mass which is added to the entire vehicle mass. The wide
validation of GEM-II against real vehicle data indicates that the
physics has been correctly implemented in GEM-II.
c) Input and output structures and how they interact with the model to obtain the expected result, i.e., fuel consumption and CO2 over
the given driving cycles;
Reviewer 1
Reviewer 2
Reviewer 3
The accompanying explanatory document emphasizes that the
input and output structure has not been finalized. A comment,
intended as guidance, is that manufacturers will be obliged to
use the executable version of GEM a large number of times to
cover each order or each truck technology change over the
year. This is needed to compute an average value for
compliance. The EPA should make every effort to insure that
the final version can be interfaced with the manufacturers'
software to insure that the process is efficient and reasonably
inexpensive, while keeping that version of the model locked to
insure compliance.
The .csv output files, viewed in Excel, provide a representation
of likely input and output files. These summaries are very
useful and appropriate.
See response for part A.
A further concern that is not discussed in the documentation is
whether any model fitting parameters were employed to obtain
the fits observed, between the GEM-derived fuel efficiency and
CO2 emissions values and the dynamometer-measured results.
The agencies are talking to various manufacturers
to make sure that our input and output structure can
be integrated with their software. The HD Phase 2
NPRM preamble seeks comments from
stakeholders on suggestions to minimize the
compliance burden associated with GEM.
Some model constants were adjusted to match the
individual vehicles, however all simulations for a
particular vehicle are run with the same
configuration.
48
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
In other words, beyond the parameters described and the
accompanying constants used in the dynamic force and energy
equations, were any other fitting techniques (or fitting
parameters) used to obtain the observed correlations between
the simulated cycle-averaged results and the SwRI chassis
dynamometer results? Presumably there were dynamometer
parameters and coefficients fitted using vehicle coast-down
data (and dynamometer operational parameters), but beyond
these, are there any other fitting parameters used to obtain the
correlations shown? Discrepancies between simulated and
measured results of less than 2-3% are probably not significant,
except in the presence of a consistent bias between the
measured and predicted, for any one vehicle, cycle or
technology considered.
Examples of adjusted model constants are
transmission shift delays, rpm limits, up and down
shifting constraints, first gear skip, etc. However
once these model constants are tuned, they are not
adjusted for individual drive cycles or tests. In
other words, vehicle-specific constants are set up to
match the behavior of the vehicle but are not tuned
for individual results. Input values directly from
experiment, such as coast down data related to aero
drag coefficient and tire rolling resistance
coefficients are taken as inputs and not further
adjusted.
Reviewer 4
Input for GEM-II:
A structure format is used to store the inputs to create the input
data for the execution of GEM-II. The structure format is
organized as follows: component.variable.units. For example,
the input variable "engine.idle_fuel_map_speed_radps"
identifies the engine speed vector of the engine idle fuel map
expressed in rad per sec. Similarly, the input variable
"transmission.clutch.input_inertia_kgm2 refers to the clutch
inertia of the transmission, expressed in kgm2. This format
follows good coding standards, making the inputs easy to pair
to the appropriate component it refers to, the particular variable
name, and the units used. Further, a modular approach is used
to store the input data for each component in separate easily
identified files in the "param_files" folder.
Although this reviewer makes excellent comments
on the model data structures based on
Matlab/Simulink, the actual code used for
certification is only an executable version of the
code, meaning that it will not require a
Matlab/Simulink license and the user would not
touch the source code, and therefore, these
comments are not relevant to the certification tool.
49
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
Output for GEM-II:
When the workspace has been populated with the input data,
the simulation model "REVSJVM" vehicle model is executed
over the user-selected drive cycle. Each major component
model (GEM_CVM, vehicle, driver, ambient) has a bus_out
output port which contains a structure of component output
data, which is used within other components. In addition,
GEM-II uses a datalog structure to store simulation output data
for later post processing to calculate emissions and fuel
economy. All the simulation output is stored in a single datalog
structure with multiple fields, each describing the component
that the data pertains to. For example,
datalog.vehicle.speed_mps refers to data log from the vehicle
component of the variable vehicle speed in m/s. This format is
an accepted coding standard within other vehicle simulation
packages (such as PS AT from ANL, RAPTOR from SwRI) as
well, making the output easy to pair to the appropriate
component it refers to, the particular variable name, and the
units used. Similarly, all variables that are used to perform an
energy balance are prefixed by "audit".
Interaction with the model to obtain the expected results:
The bus_out structure from the various components are stored
in "goto" blocks which are paired with "from" blocks to
distribute data from one block to another. The use of paired
"goto" and "from" blocks is an accepted method of decluttering
the simulation model and avoiding crisscrossing signal lines,
thereby significantly facilitating the understanding of
50
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
| information flow from one component to another.
d) Default values used for the input file, as shown in "Vehicle Simulation Model" document.
Reviewer 1
The default values are not particularly important at this stage of
development. They are sufficiently representative of recent or
current technology to provide reasonable inputs and to assess
the ability to predict real-world values. However, this is an
appropriate point to acknowledge and discuss the comparison
of GEM output with real-world measurements, as described in
the supporting material.
The real-world data serve to verify the ability of GEM to model
a variety of trucks. These data do not extend to the validation of
truck tire rolling resistance or aerodynamic drag value
selections because the validation was conducted using roller
and powertrain dynamometers, where these values were
entered and were the same values used in GEM. Generally
GEM matched the measured fuel efficiency values to within
5%, and the deviation may be attributed both to measurement
error and to the inherent simplifications in modeling. There are
some systematic errors evident in the comparison. For example,
the T700 truck efficiency is underpredicted (GEM vs.
measured) on the high speed tests and overpredicted on the low
speed tests. This trend differed for the T270 and F650 trucks.
The refuse truck fuel efficiency was uniformly overpredicted.
This leads to the conclusion that GEM may be capable of
predicting overall fuel efficiency accurately (at the 5%
difference level), but that one should still be cautious of
comparing the performance of very different technologies using
the GEM tool.
The agencies agree with this reviewer's comments.
It is a daunting task to verify every GEM sub-
model, which would require significant investment
to collect data and then validate each sub-model.
Just as this reviewer pointed out, the single overall
predicted efficiency is of interest to the
manufacturer because it predicates compliance.
GEM works reasonably well to predict this single
value, namely, fuel economy over 130 vehicle
variants.
The agencies have made significant improvements
on the driving cycle. The addition of road grade to
55 and 65 mph cruise cycles is just one example.
The agencies are seeking comment on the proposed
drive cycles and road grades in the FID Phase 2
NPRM. We will refine GEM as necessary based
on the comments for the final rulemaking.
51
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
The narrative states that "While it is encouraging that GEM
accurately simulates overall vehicle performance in an absolute
sense, it is actually more important that GEM is accurate in
relative comparisons." This is true in the sense that GEM
should encourage the best technology pathways through
comparison, but it is nevertheless the single overall predicted
efficiency that is of interest to the manufacturer because it
predicates compliance. As an example, when the axle ratios
were adjusted for a vocational truck from the chosen value of
3.76 to high (4.06) and low (3.46) values, the predicted
transient fuel economy values were 5.55, 5.49 and 5.6
respectively. In contrast to this small variation, the fuel
economy values for the 65mph operation were 7.14, 6.88 and
7.37 respectively, attributable to the 9% difference in engine
speed between the high and low ratios. This reviewer has
confidence in these relative values where a variable input is
changed. However, a comparison between two vehicles with
identical chassis and bodies, but with different engines,
transmission types and tire rolling resistance challenges several
parts of GEM in a differential sense, and a greater comparative
error must be anticipated. One cannot argue that the overall
agreement with the measured data verifies each sub-model
within GEM.
The use of single variables to represent a more complex reality
is discussed elsewhere in this review.
The cycles chosen for evaluating vehicle performance can be
52
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
changed readily in GEM. Although incorporation of grades
represents a substantial advance, the current choice of fixed
55mph and 65mph steady speeds may cause designers to "teach
to the test." [The CBD cycle, used in a previous age to quantify
transit bus performance, suffered from problems of this kind
because all steady state operation was at 20mph.] The output
data for the tractor (through engine speed / vehicle speed ratio)
show that the highest gear was used for both fixed speeds. Use
of a test cycle where speeds varied slowly through the 50 to 70
mph range could avoid pitting real world optima against model
optima, and would encourage engine downspeeding strategies
that are successful in revenue service where speed limits are
not necessarily 55 and 65mph.
Reviewer 2
See response for part A.
Reviewer 3
The default values as defined in the "Vehicle Simulation
Model" are reasonable and currently fall within the ranges of
expected vehicle values. It is not clear however under what
circumstances the user will be able or allowed to make
modifications in the final model implementation. For instance,
it is conceivable that the interplay between future auxiliary
mechanical or electrical loads on an engine might be
significantly modified through conversion of mechanical
auxiliaries to electrical or electronic devices. In that case, that
will shift the relative balance in those loads. Moreover it is not
clear that engine cooling fan loads have been adequately
accounted for in the model, as these are typically not
considered in the engine dynamometer testing from which
engine fueling maps are normally derived. This exclusion alone
can modify observed fueling rates by 10% or more under
The example mentioned by this reviewer of the
coolant fan is an example of a technology that
GEM cannot directly model at this time due to lack
of data and the extra burden of testing that would
be required to calibrate such a model. In general,
GEM will recognize all technologies that can be
measured during dynamometer cell testing through
the engine fuel mapping procedure. Those that
cannot be measured or cannot be modeled through
GEM will be accounted for in different ways
through
(1) the pull-down technologies item or
technical improvement input values
(2) Powertrain test options
(3) Innovative credit
53
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 3. Charge Question 2
Charge Question 2: Please comment on the appropriateness and completeness of the contents of the overall model structure and its
individual systems and their component models (i.e., using the MATLAB/Simulink version), if applicable, and considering the
following:
specific engine operating conditions.
Other default values including transmission gear ratios,
transmission efficiencies, axle efficiencies, tire rolling
resistance, and vehicle aerodynamic drag product seem
reasonable.
Reviewer 4
The default data used in GEM-II is complete and appropriate to
execute a simulation of HD vehicle powertrain over one of the
drive cycles available in the GEM-II default drive cycle library.
The default ambient conditions summarized in
ambient_param.m are appropriate. The default driver
parameters, summarized in driver_param.m, contain driver
gains as well as the time that the driver can look ahead in the
drive cycle. The driver gains represent an average driver, which
the user can change to emulate an aggressive driving pattern
versus a calmer driver.
The default engine maps of 270 kW, 345 kW, and 455 kW
power ratings includes inertia, idle speed. Default transmission
maps for the manual, automatic, auto-manual are also available.
Default tire radius, axle ratios, rolling resistance of the tires of
the steering axles and drive axles are also included.
The certification tool is an executable version, and
the user will not be able to access the source code
for certification. Therefore, the user will not have
flexibility to change model constants, such as
driver gain.
Table 4. Charge Question 3
Charge Question 3: When using the standard of good engineering judgment, is the program execution optimized by the chosen
methodologies?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
54
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 4. Charge Question 3
Charge Question 3: When using the standard of good engineering judgment, is the program execution optimized by the chosen
methodologies?
Reviewer
Name
Reviewer 1
Reviewer 2
Reviewer Comment
The GEM tool as presented to the reviewers provided a
workable compromise between accuracy and simplicity. It is
evident that GEM may be improved in accuracy by increasing
input data, in particular the substitution of data tables for single
values. However, these tables would need to be created for each
component at substantial expense. Also, it would be better to use
real values or tables or strategies from manufacturers, but each
of these might require audit and prescribed measurement
methods, and in many cases the manufacturer considers them to
be proprietary. The present GEM is close to being optimized:
clearly development is still taking place.
As a general observation, the engine and transmission receive
substantial attention in GEM and in much work on truck
efficiency improvement. They seem to be of interest at the
single percentage level. However, tire rolling resistance is a
major influence for vocational trucks, and drag coefficient
controls the dominant loss for freeway tractors. Yet these two
components are each represented by a single parameter. Beyond
just GEM, a more detailed consideration of these components
(e.g. longitudinal slip of tires, rolling resistance during
crosswind correction, effects of yaw on drag) would assist in
raising modeling accuracy. If that cannot be considered
expediently for this version of GEM, it should be considered in
the future, or even embedded in the tables so that it can be
applied without altering the code.
The chosen methods and execution of the model shows strong
engineering judgment throughout. A good indication of proper
EPA Response to Comment
We fully agree with this reviewer's comments. The
current version of GEM is a balance of model
fidelity and simplicity. The agencies could develop
more advanced features, but at higher cost of
development and longer duration, which would not
be able to meet the timeline of this rulemaking. In
the future, many advanced technical features
including the tire model mentioned here could be
considered.
The agencies are proposing to account for the effects
of yaw on drag in Phase 2 in the CdA input to GEM.
The HD Phase 2 NPRM preamble Section III
describes the test procedures to determine the wind
average yaw coefficient of drag value and we seek
comment on the proposed approach.
We appreciate the comments made by this reviewer.
55
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 4. Charge Question 3
Charge Question 3: When using the standard of good engineering judgment, is the program execution optimized by the chosen
methodologies?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
execution is the overall good agreement between the Phase 2
GEM simulations and testing data obtained with chassis and
power train dynamometers. The errors shown are well within +/-
5%, which is within the test-to-test variability of chassis
dynamometer testing. Execution at this level of fidelity meets
our own criteria that we have utilized to validate tractor-trailer
simulation results.5 the validation results show that the balance
between model accuracy and simplicity is adequate. As a result,
the program would be effective to model a diverse set of
technology changes and be used in regulatory applications.
References:
5 See Delgado and Lutsey (2015). Advanced tractor-trailer
efficiency technology potential in the 2020-2030 timeframe
http://www.theicct.org/us-tractor-trailer-efficiency-technology
We appreciate the comments made by this reviewer.
Reviewer:
This issue is difficult to address through the level of observation
afforded to the reviewer at this stage of development of the
model. It is not obvious that the program execution is optimized,
but the results, computational time and outputs displayed
indicate that the chosen methodologies are suitable for this
purpose.
Reviewer 4
Overall, GEM-II uses industry accepted coding practices
throughout the software modules. The following is a partial list
of these accepted practices:
• Valid variable naming structure used -
• Data bus used for each component -
• Modular components with no signals crossing -
• Useful comments to assist the user with following the
The comments made by this review are well taken in
terms of improvement of the Matlab/Simulink
version of the source code. We should point out that
the certification tool is an executable version of
GEM, and the user would not be allowed to use the
Matlab/Simulink version of the source code to
perform certification.
56
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 4. Charge Question 3
Charge Question 3: When using the standard of good engineering judgment, is the program execution optimized by the chosen
methodologies?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
code -
• Energy audit adds to the confidence level of the results
of the simulation.
• File name that is executed is echoed back to the user. If
there is a simulation abort, the debug is easier
• Data that is being loaded is echoed back to the user so
the user knows what data is being used.
• Component modules are linked to libraries. A change in
the library module propagates to all vehicle models
during execution.
At this point in the review, I do not have adequate data to
comment on the execution optimization of GEM-II. Linear
interpolation modules from the standard Simulink library are
used within GEM-II, thereby optimizing execution. If there are
any non-standard user defined functions used within the GEM-II
simulation model, the execution can be made considerably faster
through the use of s-functions within Simulink. At the time of
this review, no S-function were found in the model.
Table 5. Charge Question 4
Charge Question 4: Please comment on the clarity, completeness and accuracy of the intended output/results (CO2 emissions or fuel
efficiency output file).
Reviewer
Name
Reviewer 1
Reviewer Comment
The .csv output results are sufficiently comprehensive for the
user who is not executing the program in MATLAB for research
EPA Response to Comment
Yes, a succinct report with .csv format will be used
as output.
57
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 5. Charge Question 4
Charge Question 4: Please comment on the clarity, completeness and accuracy of the intended output/results (CO2 emissions or fuel
efficiency output file).
Reviewer
Name
Reviewer Comment
EPA Response to Comment
and design purposes. Presumably a report similar to the .csv is
anticipated as the executable version output. In fact, to the
manufacturer, who is executing this at time of sale within a
larger accounting loop, a very succinct output would be
sufficient.
Reviewer 2
The data reports did not appear to be fully complete, but the
accuracy of the output/results appears to meet reasonable
expectations. The input and output structure of GEM was not
finalized when released for peer review, however some samples
of the output files were provided to give the reviewer a flavor of
the potential structure. In my opinion, for completeness, the
output file needs to include results for each different cycle and
not only for the weighted aggregation of cycles. Some metrics
can be added to the output file to facilitate troubleshooting and
give the user a better perspective. As mentioned before, actual
simulated speeds and a measure of deviation from the speed-
distance trace would ideally also be provided for transparency of
the results. Based on the validation results, accuracy with respect
to measured data was provided and seems to be within 5%,
which is acceptable output accuracy based on comparable
modeling as well as real-world testing.
Phase 2 GEM will use the same output information
to report certification as Phase 1 GEM, meaning
that output file would only include the weighted
aggregation of cycles for the sake of certification.
However, the user does have the option to access
the source code to find out more for those
intermediate results, although the results from
source code would not be allowed for certification.
Reviewer 3
The model outputs and results seem clear, complete and
accurate. One caveat with interpreting or using the freight
efficiency or load efficiency-based results lies in the use, further
interpretation or extension of these results. Users might be
tempted to "scale" the load-based results in an inappropriate
fashion - for example, if the returned result for the computed
carbon dioxide emissions is 100 gCO2ptm (grams CO2 per ton-
For certification, GEM uses default payloads for
each regulatory subcategory, meaning that the user
would not have option to make any change on the
payload, thus avoiding the issue mentioned by this
reviewer.
58
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 5. Charge Question 4
Charge Question 4: Please comment on the clarity, completeness and accuracy of the intended output/results (CO2 emissions or fuel
efficiency output file).
Reviewer
Name
Reviewer Comment
EPA Response to Comment
mile) for say a 30 ton vehicle over a specific cycle, there might
be the temptation on the part of users to employ that same
numerical result to predict the CO2 emissions for the same
vehicle loaded to 40 tons over the same cycle. However this
assumption is not correct, as the vehicle fuel consumption is a
function not just of load or weight-related terms (rolling
resistance, grade, acceleration etc.) but also terms that are
invariant with load or weight (such as aerodynamic drag), and
this is not reflected in an emissions per ton-mile result.
Reviewer 4
The output data from a simulation execution is summarized in a
spreadsheet, which is date and time stamped, allowing the user to
verify that the output data corresponds to the simulation
executed. The output data contains data on which technology
improvement (weight reduction, vehicle speed limiter, single
drive axle, par time single drive axle, low friction axle
lubrication, predictive cruise control, high efficiency AC
compressor, electrified engine cooling pump, extended engine
idle reduction, automatic tire inflation ) was assessed, engine,
transmission type, fuel economy and CO2 emissions.
The output data, summarized above, serves the original purpose
of GEM-II, which is to enable users to demonstrate compliance
with regulatory standards either without any modifications to the
HD vehicle, or analyze the fuel economy and CO2 emissions,
when one or more technology improvements are employed.
However, information on the drive cycle is missing or not clearly
identified. In addition, key plots of vehicle tracking the drive
cycle, engine operating speed-torque points over the drive cycle,
GEM is a certification tool, and as such produces
simplified results, which can not only be easily
connected to the agencies' certification tool box for
compliance, but also can be directly linked to the
manufacturer's information technology system. For
more comprehensive outputs and plots, the user has
the option to use the Matlab/Simulink version of the
source code to provide a better understanding of the
results. GEM as a certification tool is not meant to
be used as a development aid. Again, the results
from source code would not be allowed for
certification.
59
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 5. Charge Question 4
Charge Question 4: Please comment on the clarity, completeness and accuracy of the intended output/results (CO2 emissions or fuel
efficiency output file).
Reviewer
Name
Reviewer Comment
engine efficiency contour plots, transmission operating points,
and other plots that assist OEMS to further fine tune the
powertrain and improve fuel economy/ CO2 emissions in case of
non-compliance. It would be desirable to have the results of the
energy audit summarized in the output file. The date stamp
column is appropriate for the user to cross check simulation runs.
EPA Response to Comment
Table 6. Charge Question 5
Charge Question 5: In your opinion, are there any procedures or observations that would have added to the quality of the GEM tool?
Any recommendations for specific improvements to the functioning of the outputs of the model?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
Reviewer 1
Words of caution may help in presenting GEM to the user and to
the public. It is important to state that GEM has acceptable
accuracy in predicting the fuel efficiency of a specific vehicle
under specific circumstances. However, it is equally important to
state that if GEM is used to compare two competing, but very
different, technology packages, it may not have the fidelity or
granularity to evaluate which is better. GEM may not determine
the relative difference between the two with high fidelity, and
that relative difference will depend on the vehicle vocation or test
cycle used.
Although modeling test weights were provided for the example
vehicles, there was no quantitative discussion of choice of test
weight for GEM. Test weights receive only brief mention in the
supporting documentation. It is possible that within vehicle
classes, or across regions with different topography, engine size
Test weight is set to a default value for each
regulatory subcategory, which means that the user
would not have option to make changes. This is
necessary, because allowance of different test
weights per vehicle can result in numerous standards
for the same class vehicle, which would be
challenging for the purpose of compliance. The
agencies provide details regarding the proposed
payloads and test weights in Chapter 3 of the HD
Phase 2 Draft RIA and seek comments regarding the
weights in the preamble.
60
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 6. Charge Question 5
Charge Question 5: In your opinion, are there any procedures or observations that would have added to the quality of the GEM tool?
Any recommendations for specific improvements to the functioning of the outputs of the model?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
may be selected based on the anticipated load. The distance
based strategy shows an appreciation for this issue, but some
compensation with test weight for engine size could be
considered. Perhaps test weights could be selected so that the
time to complete the route remains within reasonable bounds for
more lightly powered vehicles, but at least that highly powered
vehicles are acknowledged to be appropriate for some
occupations, loads or regions. The vocational tractor option in
Phase I is not a comprehensive solution.
Reviewer 2
Phase 2 GEM could generate two different output reports. One
that only includes the most relevant information on an aggregated
format and is used only for compliance purposes, and a second
one, that is very detailed and includes results disaggregated by
cycle and other relevant information that may help the users to
troubleshoot their results, learn the inner workings of the model
and potentially suggests enhancements to it. I suggest including a
summary of the energy audit in the output. Also, provide the
average engine efficiency over the cycle for the different test
cycles, as well as the ratio of average engine efficiency to
maximum engine efficiency, which is an indication of how well
the transmission parameters are matched to keep the engine
operating near its peak efficiency range.
The output file provides some basic "sanity checks" such as
number of shifts, ratio of number of shifts to number of gears in
the transmission, distance traveled, and ratio of actual time to
target time. Please also provide ranges of valid or acceptable
values for these parameters so the user can be aware of any
GEM is a certification tool, and as such produces
simplified results, which can not only be easily
connected to the agencies' certification tool box for
compliance, but also can be directly linked to the
manufacturer information technology system. For
more comprehensive outputs and plots, the user has
the option to use the Matlab/Simulink version of the
source code to provide a better understanding of the
results. GEM as a certification tool is not meant to
be used as a development aid. Again, the results
from source code would not be allowed for
certification.
61
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 6. Charge Question 5
Charge Question 5: In your opinion, are there any procedures or observations that would have added to the quality of the GEM tool?
Any recommendations for specific improvements to the functioning of the outputs of the model?
Reviewer
Name
Reviewer Comment
potential issues with the simulation.
A further step that could allow the tool to be much more useful
would be to allow users to input their own cycles as is currently
done with VECTO6 (Vehicle Energy Consumption calculation
Tool) model in Europe. The VECTO tool has a "declaration
mode" for compliance, and an "engineering mode" which offers
the ability to edit inputs and allow users to explore what the tool
can do. This would be critical for transparency and follow the
best practice as seen in the Europe situation. It would also be
highly useful for individuals in the heavy-duty vehicle supply
chain to explore the variation of the results with respect to real-
world duty cycle factors. Especially considering the very diverse
use of heavy-duty vehicles in local, regional, and long- distance
conditions, this capability would allow dealers and fleet
managers to gauge how fuel consumptions for particular relevant
driving patterns differs from the cycle. This would help ensure
the technologies that are more suited to particular duty cycles are
being selected in the market place, and it would also help
overcome the prevailing market barrier, whereby knowledge,
data, and confidence on truck efficiency has been limited.7
References:
6SeeLuzetal(2014)
http://ec.europa.eu/clima/policies/transport/vehicles/heavv/docs/f
inal report co2 hdv en.pdf
7 See Roeth et al (2015) http://www.theicct.org/hdv-technolosv-
market-barriers-north-am erica
EPA Response to Comment
62
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 6. Charge Question 5
Charge Question 5: In your opinion, are there any procedures or observations that would have added to the quality of the GEM tool?
Any recommendations for specific improvements to the functioning of the outputs of the model?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
Reviewer:
The version reviewed here does not include the graphical user
interface (with "pull-down menus") described in the instructions.
The ability to modify input parameters and vehicle attributes will
improve the user experience, while obviously presumably not
impacting the model outputs.
Due care and attention should be paid to the number of
significant digits presented in the output results. For example, in
the results presented from a single specific simulation (shown
below):
» GEM_Phase2_Idle_55_65_CARB_HHDDT_Transient
Distance = 32.414 mi
Fuel Consumption = 5.4059 gallons
Fuel Consumption = 17184.1217 grams
Fuel Economy = 5.996 mpg
Fuel Consumption = 530.140 g/mile
CO2 Emission = 1697.77 g/mile
The number of significant digits in the model simulation outputs
presented above (some of which are directly related through
derivation or calculation) varies from 4 to 9. This does not meet
recommended practices in the presentation of results and data.
Moreover, industry experience in the measurement of real-world
fuel efficiency during over the road truck testing dictates that
measured fuel consumption variations of less than 1-2% should
not be considered significant or compelling, and this level of
variation could correspond to variations in the 2nd or 3rd
significant figures in fuel consumption in most cases.
The agencies have consulted with all major
manufactures on whether a graphic user interface
(GUI) needs to be developed. At this time, none of
them support this idea. The reason is that a GUI
would prevent the GEM output from being
integrated with their internal information technology
system. At the same time, a GUI would provide
convenience to those individual user or small
entities for use. In order to compromise this issue,
the agencies decide not to propose a GUI. Rather,
the input file is designed in such a simple way that
the individual users can easily use it for their own
purpose, and an interactive Excel spreadsheet will
be made available as mentioned previously.
The number of the significant digits shown in the
output screen is not the same as those in certification
output file (.csv). It should also be noted that the
same output format is used for light duty vehicles
where the magnitude of the numbers is much
smaller and therefore the significant digits are fewer.
No attempt was made to tune this output, which is
not part of the certification process, for heavy duty
test results.
63
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 6. Charge Question 5
Charge Question 5: In your opinion, are there any procedures or observations that would have added to the quality of the GEM tool?
Any recommendations for specific improvements to the functioning of the outputs of the model?
Reviewer
Name
Reviewer Comment
EPA Response to Comment
Reviewer 4
The following modules will enhance GEM-II:
• A module that is able to create the input data for a GEM-
II execution from a user provided spreadsheet with pre-
defined tabs for the engine, transmission, drive cycle,
vehicle parameters, and technology improvement. Users
are more familiar with spreadsheets than the Matlab
environment.
• A GUI module that guides the user to create a HD vehicle
model.
• A module that allows users to select plots of key
component performance. These plots may be summarized
in the output spreadsheet data file.
• A more detailed explanation of all the user provided data
"target.X" and the various choices available for each of
the user provided data. For example, what are the choices
that the user has for the variable target.veh_style ?
• GEM-II execution takes place within the Simulink
environment. During execution, no feedback is provided
to the user on the status of the simulation. The user is
waiting on a blank screen - Percent complete of the
simulation and which drive cycle is being executed would
be useful feedback for the user.
• Predefined sample input data for all class of vehicles to
assist users to easily modify them if necessary, since
some users may not be familiar with Matlab.
• The ability to turn on the feed forward term for the driver
model in case of tracking problems in drive cycles with
grade.
Many comments made by this reviewer are very
helpful in terms of the Matlab/Simulink version of
GEM. As pointed out earlier, only the executable
version of GEM is allowed for certification where
the Matlab/Simulink environment is not needed.
Because of that, the user has no option to make any
model constant modifications.
A GUI is helpful for individual users. However, it
may not be helpful for large vehicle manufacturers
where they need to connect their information
technology system with the output of GEM. The
agencies attempt to balance the needs of individual
users and large organizations. The combination of
the proposed input file and GEM User Guide should
be as user friendly as a GUI.
64
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 6. Charge Question 5
Charge Question 5: In your opinion, are there any procedures or observations that would have added to the quality of the GEM
Any recommendations for specific improvements to the functioning of the outputs of the model?
Reviewer
Name
Reviewer Comment
• The ability to model accessory power draw as a function
of engine speed and engine temperature. The engine
cooling fan power cycle will affect fuel economy and
CO2 emissions.
tool?
EPA Response to Comment
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Reviewer
Name
Reviewer 1
Page
8
8
10
16
16
Paragraph
1
2
2
2
2
Reviewer Comment
Axle lookup table is a better approach
"Brakes" Addition of the inertia component to axle
is curious/ a curious description. This is rather a
retarding torque. Perhaps the inertia becomes a force
if MATLAB is viewed over time steps. I was
confused by this language.
It would be good to see (as well) the equivalent to
Figure 1-3 with the proposed locked shifting
strategies in GEM, rather than actual strategies as
discussed on p. 10.
Need to be cautious about claims in modeling
changes. Effect of change in Crr is clear, but GEM
may have difficulty with relative accuracy in
changing transmission type.
Note in presenting these agreements that Crr or Cd
changes are entered as dynamometer A-B-C
coefficients, and these are not real world
EPA Response to Comment
The agencies note the comment but
determined no revisions are required to
the documentation
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM.
Some of the comparisons are shown in
Table 4.
The agencies note the comment but
determined no revisions are required to
the documentation
The agencies note the comment but
determined no revisions are required to
the documentation
65
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Reviewer
Name
Reviewer 2
Page
17
18
19
22
23
2
2
2
Paragraph
Table 1-3
2
Fig 1-9 to
1-13
Table 1-4
End of
page
2
2
3
Reviewer Comment
measurements. Essentially the dynamometer is
partially modeling these effects.
And the figure. Note that the 1.8% error represents a
17% error based on the difference, if this figure is
intended to show the accuracy of predicting
differences rather than absolutes.
GEM is using automated shifting, essentially, for
both MT and AMT, and this will be used (with PI
control pedal) for impact assessment. Yet this
paragraph brings human drivers to the fore.
Philosophically, that the human drivers are the truth
and the accuracy of revenue service, more than the
model or powertrain cell.
Are these with default or manufacturer's shift tables?
Little or no shifting occurs in the 55 and 65 cycles.
That should be stated.
Text should explain in a little more detail how these
powertrain tests may be inserted into GEM
The list of key technical features may include the
fact that the new model uses distance-based cycles
instead of time-based cycles, and the fact that test
cycles now include road grade.
The claim "more stable engine idle speed controller"
is not discussed in the text. Some metric or
quantification of what is meant by "more stable"
needs to be provided.
Regarding "substantial effort has been put forth to
EPA Response to Comment
The agencies note the comment but
determined no revisions are required to
the documentation
The agencies note the comment but
determined no revisions are required to
the documentation
Yes.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The agencies note the comment but
determined no revisions are required to
the documentation
It is a distance-base model
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
It is mainly used for internal use.
66
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
accurately track and audit power flows through the
model to ensure conservation of energy" The report
lack details about the energy audit. Was the energy
audit developed just for internal quality control or is
it going to be provided to the end users in an output
file?
Regarding "the road gradient has been modified to
accept a road grade that varies as a function of
distance traveled" Please introduce the concept of
distance-based versus any pros/cons of the new
method and why did you change the approach.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
Driver subsystem. This section (1.2.2.2) is not clear
to the reader and, in my opinion, needs rewriting.
There are various confusing statements such as "the
feed forward calculations using drive cycle
accelerations and vehicle mass have been removed".
The section also mentions (page 4, paragraph 1) that
a ratio of speeds (which is non-dimensional) is
integrated to produce the current cycle position
(which has units of distance), which is dimensionally
incorrect. I recommend showing the equations to
avoid confusing the reader. Regarding the statement
"the addition of distance compensation allows all
simulated vehicles to complete an equivalent trip
such as traveling from point A to point B" Does that
mean that without distance compensation the
vehicles would not complete the trip?
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
How were the proportional and integral constants of
the PI controller estimated? Is the same driver
subsystem used independently of transmission
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
67
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
choice? How was the look-ahead feed-forward
control implemented?
Consider removing the mention to the "variant
power train architecture" since it is not mentioned
anywhere else in the report.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
Please clarify that the engine map is not a pre-
determined parameter as in Phase 1 GEM, but a
user-defined input.
For consistency with previous section, please
show the proposed constant power loss magnitude
of electric subsystem.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
Please change "four different variants" for "three
different variants" (MT, AMT, AT).
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
11
There is an apparent incongruence in Page 5,
paragraph 4, which mentions that each transmission
"features a unique control algorithm matching
behaviors observed during vehicle testing"
however, in Page 5, paragraph 5 it is mentioned
that "all of the transmission models use an auto
shift algorithm to determine the operating gear over
the cycle". Are the auto shift algorithm parameters
changed based on transmission type? Is there
really a unique control algorithm for each
transmission type?
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The "auto shifting optimizer" needs proper
documentation. How does it work?
A reference paper (Newman, K., Kargul,}.,
and Barba, D., "Development and Testing
of an Automatic Transmission Shift
Schedule Algorithm for Vehicle
Simulation," SAE Int. J. Engines
68
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
12
8(3):2015, 2015. doi:10.4271/2015-01-
1142.)
has been added into Draft RIA chapter 4 for
more information
The clutch model is a key new addition and lacks
proper documentation.
Please support claims such as "realistic actuation
durations and more accurate physics of torque
conservation and lockup behavior" with data.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
Please clarify what you mean by "This layout is
more similar to a manual transmission, but the
application for a planetary gearbox is a
reasonable approximation as this type of gearbox
can utilize a variety of topologies" i s confusing to
me.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The statement "The brake model also adds a
rotational inertia component to the axle" is
misleading since the inertia of brakes is set to zero
in GEM.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
Please provide a table with a complete list of pre-
defined and user-defined parameters. There is no
need for specific values, just the list of parameters.
It is included in the NPRM
Description of test condition number 6 reads,
"Run a new set of road load coefficients to
represent a vehicle configuration optimized for
fuel efficiency for each vehicle that was tested."
To be consistent with the other test conditions
described please quantify the reductions in rolling
resistance, aerodynamic drag, and mass for this
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
69
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
particular "optimized package" case.
13
Figure 1-3
to 1-6
In the legend, it is not clear if the 55mph and 65
mph tests contain grade or not. Also, the "utility"
cycle is not described in the text.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
15
Figure 1-6
Although is mentioned in the text, it seems that
the refuse truck was not tested under the refuse
cycle. Also, it seems that it was not tested under
different test conditions as the remaining vehicles.
Any reason for this? Please explain.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
16
18
18
Change "numerically" for "numerical."
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The statement "In some cases, it is hard to
quantify which method, vehicle simulation or
chassis dynamometer test, is more accurate" is very
misleading. The chassis dynamometer test
although imperfect, is a measurement and it was
treated as a "true" value in the validation process
(Figures 1-3, 1-5, and 1-6). Accuracy can only be
measured in reference to a true value. The chassis
dynamometer test was selected as such and the
model cannot hope to have better accuracy than
the test.
Added SwRI technical research workshop
as one of the references to Draft RIA
Chapter 4. (US EPA, "Technical Research
Workshop supporting EPA and NHTSA
Phase 2 Standards for MD/HD
Greenhouse Gas and Fuel Efficiency —
December 10 and 11, 2014,"
http://www.epa.gov/otaq/climate/regs-
heaw-dutv.html
In the statement "GEM is capable of capturing the
impact on the total vehicle CO2 emissions and
fuel consumption due to any technology
improvement" the word "any" is misleading since
the validation effort was only done over
aerodynamic, rolling resistance and mass
parameters. Moreover, there is certainly a set of
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
70
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
technologies that GEM is incapable of model.
17
Table 1-3
Correct the numbers on the "Delta" column. Due
to rounding it is not evident that the column
represents the difference between the two previous
columns. For example in the fourth row we have
3.9% - 4.9% = -0.9%. In the same Table, please
calculate the relative error (consistent with Figure
1-8). Is this relative error more relevant than the
"delta"? What is the maximum acceptable relative
error?
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
22
Please add a table that summarizes the
technologies captured by the simulation and the
technologies that are not captured but are
accounted for via drop-down menus. A third
(optional) column may include technologies that
currently are not either simulated, nor recognized
by drop-down menus but potentially may be
included in future regulations.
We have detailed the technologies included
as inputs to GEM and technology
improvement inputs (drop-down) in the HD
Phase 2 NPRM preamble. We welcome
comment on other technologies that would
only be considered as part of the innovative
technology credit program and therefore not
simulated or recognized by the drop-down
menu.
22
Section
1.4.1.1
Regarding "As described in Chapter 1.2.2, one of
the major changes in the HD Phase 2 version of
GEM is to allow manufacturers to enter their
transmission gear number versus gear ratio"
Chapter 1.2.2 does not mention that fact. Please
check throughout the report to avoid issues with
these self-references.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
23
Regarding the statement "Manufacturers also have
an option to select the type of transmission, which
is either manual or automatic" It is not clear if
transmission type is a required input or an
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
71
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
optional input. Also, the user should have three
options (AT, MT, and AMT), not just two.
23
Table 1-4
Table heading should mention that the results are
from simulations in GEM, not measurements. Also,
it is not clear if the 55mph and 65mph cruise contain
grade.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
23
The report mentions that due to lack of data, DCT,
DCT with TC and Allison TC-10 AT transmissions
were not validated. That assertion implies that MT,
AMT and AT were validated. Since transmission
technologies were not tested at the same level of
detail as road load reduction technologies, it is
important to acknowledge the different level of
validation between transmissions and road-load
reduction technologies.
The agencies note the comment but
determined no revisions are required to
the documentation
23
Last
Paragraph
Regarding DCT and other transmission types not
included in the model, the report says, "The
manufacturers still have the options to use
powertrain dyno tests to quantify the benefits of
these or any other special transmissions". It is not
clear if the results of the power train tests are going
to be used to correct the GEM simulation results or
are going to replace the GEM simulation results
altogether. Please clarify.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
24
Regarding OEM overriding the axle efficiency
input, "the inputs would be determined by using
the prescribed test procedure" It is not clear
which test procedure is the report referring to.
Please add a reference to such procedure.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
24
Last
Rolling resistance coefficients are usually expressed
The text has been corrected (or clarified]
72
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
Paragraph
25
25
25
in units of kilograms per metric ton (11 = 1,000 kg).
The units used (kg/ton) may imply short ton (1 ton =
2,000 Ibs.). Please correct and be consistent
throughout the document.
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The "Regional", "Multi-purpose", and "Urban"
composite duty cycles are not described in the
document.
More can be seen in Chapter 3.4.2.3 about
idle cycle as stated in the same paragraph
Please support your statement, "We concluded that
for the 55 mph and 65 mph duty cycles, OEM's
interpolation of steady-state data tables was
sufficiently accurate versus the measured results"
What was the observed accuracy of using steady-
state maps for the 55 and 65mph cruise cycles?
How much accuracy is sufficient?
This statement is based on the engineering
judgment.
The 55mph cruise is named "urban highway with
road grade", the 65mph cruise is named "rural
highway with road grade, and the ARE transient
cycle is named "urban local". Please try to use
consistent names throughout. Also, since the
cruise cycles in Phase 1 were time-based and did
not include road grade, the Phase 2 cruise routes
(distance-based with grade) are not equivalent to
them and using the same name is misleading. I
think a distinction can be made by using the term
"route" for the distance-based tests and "cycle"
for the time-based, or simply name them "55mph
cruise with grade" and "65mph cruise with grade".
The agencies note the comment but
determined no revisions are required to
the documentation
29-
30
Table 1-6
&1-7
Automatic transmissions have the same efficiency
for all gears (98%), and such efficiency is
equivalent to that of manual and automated manual
We chose this based on the comments from
manufacturers' input plus our engineering
judgment.
73
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Reviewer
Name
Page
Paragraph
Reviewer Comment
EPA Response to Comment
transmissions in direct drive. Was this advantage
observed during testing? If possible, please
document with measured data showing the benefits
of AT over MT in terms of gearbox efficiency.
Reviewer 3
Use "dynamometer" and not "dyno"
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
Et
seq.
Use "Phase 2" and not "Phase II"
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
1.2.2.3.1
Use "watts" or "W" and not "Watts"
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
1.2.2.3.1
"it may o make use" requires clarification
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
1.2.2.3.3.7
"With the new gear engaged the clutch is reengaged
and the engine is again allowed to operate at full
load." This statement presupposes that the
transmission was shifting under full engine load,
which is not necessarily the case for high power
engines under benign operating cycles.
The agencies note the comment but
determined no revisions are required to
the documentation
1.2.2.3.4
"This includes drive shafts as well as driven and
passive axles, consisting of a differential, brakes and
tires." Passive axles will not ordinarily include a
differential, only driven axles will.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
1.2.2.4
"The vehicle system consists of the chassis, its mass
and forces associated with aerodynamic drag and
changes in road grade". Why "changes in road
grade"? Any constant road grade (other than zero)
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
74
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Reviewer
Name
Page
9
9
10
11
12
13
20
21
22
Paragraph
1.2.2.4
1.2.3.2
1.3
1.3.2
Fig 1-11
Fig 1-13
Reviewer Comment
will have an effect on the apparent vehicle load, and
not just changes in grade.
"computes acceleration [not accelerations] from the
input force and equivalent mass which is integrated
to generate vehicle speed and distance traveled"
Use "Matlab" or "MATLAB" and not both.
"Validations use all actual vehicle variables
conducted at Chassis dyno cell," needs editing.
Ref. 3 is not the SwRI report as stated. It is an
ASME Technical Paper.
A +-15% variation in aerodynamic drag, for
example, is unlikely to span the full range of
expected values in the future for HD vehicles.
Presumably aerodynamic modifications to Class 8
vehicles may result in significantly lower drag
coefficients, or drag products (CdA)
For the Class 8 T700 tests, at high fuel efficiency,
the GEM model appears to consistently under-
predict the actual measured vehicle fuel economy.
This consistent offset is of concern. The reverse is
observed for the Class 6 truck tests.
Caption is incorrect.
Vertical axis incorrect unit designation.
Do not use the term "aero drag". It should be
"aerodynamic drag".
EPA Response to Comment
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The Technical Paper is the report
This was just for the purposes of the tech
study, these are not GEM limitations
In the technical research workshop held at
SwRI in December, 2015, the testing
variability is discussed in details. It is very
challenging to validate GEM for those
highly transient cycles for a heavy truck.
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
It is NM for torque
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
75
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 7. Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Reviewer
Name
Reviewer 4
Page
22
5
3
6
9
Paragraph
2
2
4
3
Reviewer Comment
"tire radius" should be "tire rolling radius" or
'effective radius'.
The last sentence reads "If a manufacturer uses a
hybrid powertrain for the power take-off devices, it
may o make use of ... .". The "o" after may is a typo.
Distance compensation is critical for all vehicle
simulations - Therefore, this is a good feature that
has been implemented in GEM-II
Please explain what is included in spin losses, since
this may not be clear to all OEM users.
The GEM-II executable is very appropriate for users
who are not fluent in the Matlab/Simulink
environment. Further, the executable prevents users
from making any changes to GEM-II to support
compliance.
EPA Response to Comment
Phase 2 NPRM
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The text has been corrected (or clarified]
in Chapter 4 of the Draft RIA to the HD
Phase 2 NPRM
The agencies note the comment but
determined no revisions are required to
the documentation
The agencies note the comment but
determined no revisions are required to
the documentation
The agencies note the comment but
determined no revisions are required to
the documentation
Table 8. Specific Observations on Electronic Model Entitled, "GEM Tool"
Reviewer
Name
Reviewer 1
Reviewer 2
Input
Variable
Reviewer Comment
No Specific Comments or Observations
Provided.
File: "GEM run_postproc.m" Lines: 104 to
108. Potential issue: It seems to me that the
equation is not dimensionally correct (I
might be wrong). Additive terms should
have the same units and the equation seems
EPA Response to Comment
The equation, while lengthy, is correct. Simplified,
the equation amounts to grams/hour * hour/mile
which equates to grams/mile.
76
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 8. Specific Observations on Electronic Model Entitled, "GEM Tool'
Reviewer
Name
Input
Variable
Reviewer Comment
EPA Response to Comment
to be adding gCO2/h terms with gCO2/mile
terms. Please check for missing terms
and/or appropriate use of parenthesis.
File: "GEM_run_postproc.m" Line: 101.
The calculation performed for case 4 is
identical to the calculation performed for
case 3, making it redundant. Please
eliminate case 4 (neutral idle with
start/stop), or correct the equation to
account for a 90% reduction in neutral idle
emissions (not drive idle emissions).
This is correct, for certification neutral idle and
start/stop will not be credited simultaneously, only
the start/stop credit is available. This code is from
the evaluation phase of GEM development.
"vehicle.chas
sis.
frontal_area_
m2"
The name is misleading. Please rename to
"drag_area". Drag area is CdA,
aerodynamic drag coefficient is Cd and
frontal area is A.
The model supports separate variables for Cd and
Area. For certification, we set the Cd to 1.0 and
apply the Cd*A to the area variable. This does not
affect simulation results.
File: "load_GEM_params.m" Lines: 76 to
85. These equations set weight reduction
penalties for spark-ignited CNG (525 Ibs),
compression-ignited CNG (900 Ibs), and
compression-ignited LNG (600 Ibs). Please
set a weight reduction penalty for the
missing case: spark-ignited LNG.
We are seeking comment on the weight penalties
associated with natural gas vehicles in the preamble
to the HD Phase 2 NPRM. We will adjust the
values accordingly for the final version of GEM,
based on the comments received.
File: "load_GEM_params.m" Line: 94.1
am confused about the variable
"vehicle.chassis.mass_dynamic_kg" as
defined here is a constant. However, during
the simulation (variable:
datalog.vehicle.dynamic_mass_kg)the
equivalent mass of the rotational
components vary depending on the active
The chassis.dynamic_mass_kg represents the weight
of the chassis ("static_mass_kg") plus the effective
mass of the rotating inertia of the tires/wheels (125
Ibs mass per wheel as consistent with other EPA test
procedures). The upstream inertias are calculated
dynamically during model execution, as noted, and
added to the chassis.dynamic_mass_kg to determine
the vehicle's total effective inertia. The datalog
77
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 8. Specific Observations on Electronic Model Entitled, "GEM Tool'
Reviewer
Name
Input
Variable
Reviewer Comment
EPA Response to Comment
gear.
variable represents the total vehicle inertia during
simulation, not just the inertia of the chassis.
File: "load_GEM_params.m" Lines: 117
and 143. It is not clear what exactly the
variables
"transmission.autoshift.cost_map" and
transmission.autoshift.required_cost_ben
efit_ratio" represent, and how are they
used in the model. Other features such as
"restrict skip shifts", restrict shift parity",
"disable coast saving downshifts" are not
well documented in the .m files or in the
report.
These features are not user inputs but are described
in a technical paper, SAE 2015-01-1142.
File: "load_GEM_params.m" Line: 131.
The calculation of
"Transmission, gear. inertia_kgm2"
involves a multiplication by 0. Therefore
it would produce an array of zeroes. Also,
assuming that the value of 0 is changed
to a finite constant, it seems that the
model is assuming the same inertia value
for all the different gears. Is that
simplification accurate?
For the transmission in question, the total inertia is
represented by the common input and common
output inertia, the inertias referred to are the "gear
specific" inertias which for an AMT or MT are set
to zero since all gears always spin with the input or
output. Since this is not a user input the default
value makes little difference for certification
purposes since all users will use the same value.
File: "load_GEM_params.m" Line: 134.
The calculation of
"transmission.gear.spin_loss_torque_Nm"
involves a division by a factor of 3.73 for
all the transmissions that are not
C78_AMT. I am wondering what that
factor represents. Also, if that factor is still
The 3.73 is a scale factor which scales the default
Class 8 spin losses down to a level comparable to
the automatic transmission while maintaining the
same loss trend as the Class 8 AMT with respect to
speed. The default transmission losses will receive
additional scrutiny between now and the final rule.
In addition the peer review does not necessarily
78
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 8. Specific Observations on Electronic Model Entitled, "GEM Tool"
Reviewer
Name
Input
Variable
Reviewer Comment
valid for a potential C78 MT transmission.
Energy Audit. Net system kinetic energy
change is 0 kJ, which is the result of the
test cycle starting and ending at 0 mph
speed. The test cycle seems to start and end
at the same altitude (symmetric grade
profile traveled at constant speed), so one
can expect the net system potential energy
change to be OkJ as well. Since both kinetic
and potential energy are conservative and
not dissipative forms of energy, I was
confused about the energy audit accounting
for energy "consumed" by gradient at about
17% of the losses for a tractor-trailer.
Checking the equations, it seems that the
energy audit is performed only for positive
tractive loads. If that is the case I am
confused about the energy consumed by the
brakes at about 7% of the losses for a
tractor-trailer. They should be really low if
one is only considering positive tractive
loads. The brakes are mostly applied for
negative tractive loads (e.g. driving
downhill). Am I missing something here?
Why the weighted average speed
(weighted avg speed mph) used in post
processing (file: "GEM_run_postproc.m")
is based on the target speed-distance trace
and not in the actual simulated speed-
EPA Response to Comment
have specific default values for every transmission
type by capacity (LD versus HHD, etc).
For auditing purposes, the potential energy changes
are split between "Energy Consumed" (going uphill)
and "Energy Provided" (going downhill). In the
case of a driving cycle which starts and ends at the
same altitude but has some gradient in-between you
will see equal (within model tolerances) entries in
both of these audit categories.
All of the "Energy Consumed" audit values
represent energy sinks, the brakes are included
among these since they dissipate kinetic energy.
In any case, the audit report is for internal use only
and does not represent a user or certification output.
For tractor-trailers the idle weight is zero and the
simulation grams/mile are multiplied by target mph
and also divided by target mph so what remains is
simulation grams/mile which will reflect the
modeled performance (or under-performance) of the
79
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 8. Specific Observations on Electronic Model Entitled, "GEM Tool"
Reviewer
Name
Reviewer 3
Input
Variable
Aerodynamic
drag
Engine
Accessories
Reviewer Comment
distance trace? This decision has
implications for underpowered vehicles
that deviate substantially from the target
trace since their actual average speed may
differ substantially from the target speed.
This also has implications for the
calculation of idle fuel rates for vocational
vehicles since the conversion factor from
idle fuel rate in [kg/h] to [kg/mile] is the
weighted average speed.
- Audit data for
GEM_Phase2_Idle_5 5_65_C ARB_HHDD
T Transient drive cycle -
"Energy Consumed by Cd" should refer to
"CdA" and not "Cd" alone.
Engine Accessories = 846.16kJ 0.23%
It seems as though engine accessory loads
are under-accounted for in this
implementation of the model - a 0.23%
loss for a full transient cycle seems
inappropriately low.
Usable System Energy Provided =
373988.49 kJ
Engine Energy = 309402.35 kJ
Engine Efficiency = 42.08 %
EPA Response to Comment
vehicle.
For vocational vehicles the same is true with regard
to the simulation grams/mile over the drive cycles.
Idle consumption takes places at zero speed and is
measured in grams/hour so there is a conversion
factor required to obtain grams/mile. The target
weighted average speed represents that conversion
factor and does not alter the modeled vehicle
performance (or under-performance) over the drive
cycles.
True, but the audit report is for internal use only and
does not represent a user or certification output so
the shorthand here is of little importance.
For peer review, the accessory load is set to 1300 W
combined mechanical and electric load for all
vehicle classes. The agencies are seeking comment
on the predefined values in GEM in the HD Phase 2
NPRM preamble and will make the necessary
adjustments for the final version of GEM.
The maps provided were mathematically generated
and do not necessarily represent actual fuel maps.
In any case this would represent the total efficiency
over the drive cycle. These audit results are not user
or certification outputs and will not be available
80
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Table 8. Specific Observations on Electronic Model Entitled, "GEM Tool"
Reviewer
Name
Reviewer 4
Input
Variable
Reviewer Comment
This integrated "engine efficiency" is high
for the average efficiency expected across a
fully transient cycle - does this refer to the
peak engine efficiency encountered?
Avoid taking the derivative in the Simulink
models. This can cause instabilities if the
signal fluctuates rapidly
EPA Response to Comment
during certification.
Point taken, the use of derivative blocks is
minimized in GEM.
81
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
E. INDIVIDUAL PEER REVIEWER COMMENTS
82
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Reviewer # 1
83
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Review Comments on EPA's Heavy-Duty Greenhouse Gas Emission
Model (GEM): Phase II and Supporting Documentation
I. GENERAL IMPRESSIONS
The purpose of GEM in determining standards, providing a compliance tool, and estimating real-
world benefits is clearly articulated. The proposed Phase II GEM tool has evolved substantially
from the Phase I version, particularly by allowing additional user inputs that are necessary to
acknowledge fuel efficiency design improvements. As a compliance tool GEM should seek to
account for all technology advances and differences, yet remain simple to execute and employ
readily-measured and well-defined input variables. The Phase II GEM model has reached a
reasonable compromise between these conflicting goals, and this made clear in the narrative. The
overall architecture is sound, the present component models are appropriate to the task in nearly
all cases, and they are clearly described. The model yields credible results and credible responses
to input variable changes. The overall model predicts fuel efficiency to within 5% of
experimental measurements, and a clear summary of these comparisons is presented. However,
the material does not address quantitatively the experimental error that is likely in these
comparisons, and measurement errors associated with rolling resistance and drag are not
included in the data comparison. Overall agreement with experimental data (as vehicle fuel
economy) does not validate each component model to the same degree, and this should be
acknowledged more clearly. As one example, overall fuel economy data are used to compare a
component model with a fixed efficiency value. The difference overall was only 1.67%, but this
represented more than a third of the losses for that component. The move to a distance-based
strategy is justified and well-described, and represents a laudable advance. Addition of several
transmission models is presented, as are thoughts on the addition of transient adjustment factors.
Little is said of the "powertrain variant architecture," although this may prove important for
integrated powertrains with or without a hybrid component. Transient Adjustment Factors are
not yet finalized, and transient operation may warrant more than a single correction factor
approach.
II. RESPONSE TO CHARGE QUESTIONS
1. Please comment on EPA's overall approach to the stated purpose of the model (meet
agencies' compliance requirements) and whether the particular attributes found in the
resulting model embodies that purpose. Were there critical results or issues that were not
discussed or addressed by the GEM tool or its component sections?
Throughout his review, except where different fuels are mentioned, fuel efficiency improvement
and GHG reduction are considered to be synonymous.
Beyond the need for assurance of compliance is the need to reduce fuel use and climate change
emissions from heavy-duty vehicles in revenue use. A successful EPA approach must be
examined from two perspectives. First, it is necessary that the approach yields well-defined,
unambiguous results to allow a manufacturer to compare against a standard without the process
being unreasonably onerous. Second, it is necessary that that the vehicle attributes and behaviors
embodied in the certification process are substantially representative of real world circumstances.
These two necessities are often in conflict, because the real world scenario is complex, variable,
84
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
and long in duration, whereas the standard must be concise and precise. The conflict is far
greater in the heavy duty trucking arena than for light-duty vehicles or rail because truck
architectures vary widely, and are used in an even wider fashion.
Employing a model such as GEM to assure compliance provides some relief in the conflict
described above because the model can be executed for a variety of activities and scenarios
without excessive cost of time and resources. However, as the model is challenged to predict
these varied (and emerging) scenarios with fidelity, the model complexity rises. As a result, the
empirical tables or computational sub-models needed within the model grow in number and
demand substantial engineering time to prepare and verify. One might argue that "the more one
models, the more one measures," recognizing too that precise test protocols must accompany
each measurement.
Compromise is necessary between four fundamental needs:
1) Relevance of the model to real world truck operation.
Else the real world improvements will not match the changes in standards.
2) Accuracy of the model in predicting measured fuel efficiency (and GHG production).
Else confidence in the model will be lost and the compliance will become artificial.
3) Accuracy of the model in predicting differential effects of technology changes on fuel
efficiency (and GHG production).
Else the drivers for technology advancement will be lost
4) Control complexity and cost of modeling and compliance.
Else the cost transferred to the consumers will be inappropriate.
The overall new GEM approach shows awareness of these necessities, and reaches a reasonable
compromise. However, when some parameters are fixed by the agency, manufacturers may be
discouraged from pursuing certain development opportunities in the following way. Both future
fuel pricing and future fuel efficiency standards are unknown. If high fuel prices transcend the
standards, then manufacturers will pursue every cost-effective tool to reduce fuel use. However,
if fuel prices are low and are not the driving force, and the GEM approach offers default values
for factors such as engine transient adjustment factors or transmission efficiencies, some
opportunities for real world reduction may be left on the table. GEM cannot rely on economic
drivers to address technology advances that are not modeled.
GEM is a vehicle-based tool that is geared towards road-load demands rather than engine-
specific ("%load" and "%speed") demands. Only the new distance-based cycles give a nod
toward engine power. Criteria pollutants are still characterized using a paradigm based on engine
rated output. Allowing different measurement methodologies for efficiency and criteria pollutant
production will open a window for separate hardware and software optimization for each test.
This may demean the benefits of the separate standards to some degree: in-use compliance for
criteria pollutant measurement will not close this gap because measurement allowance for
criteria pollutants reduces stringency for criteria pollutants.
2. Please comment on the appropriateness and completeness of the contents of the overall
model structure and its individual systems and their component models (i.e., using the
MATLAB/Simulink version), if applicable, and considering the following:
85
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
a) Elements in each system used to describe different vehicle categories;
The proposed GEM Phase II model represents a substantial advance over the model used to
implement the first phase of truck efficiency legislation, and encourages more technology
advances from manufacturers in consequence. Improvements in truck efficiency are based
primarily on reductions in aerodynamic drag, tire rolling resistance and engine brake specific
fuel consumption, and this was recognized in the first GEM model. Practically, there is less to be
gained from aerodynamic improvements in most vocational truck operation than in long haul
trucking and the GEM model as presented neglects vocational truck aerodynamics, and the
modelers are right to exclude aerodynamic parameter entries for low speed trucks. However, the
overall GEM structure is capable of modeling aerodynamic improvements for niche vocational
designs, and has the flexibility to extend beyond the present exclusion of the drag coefficient. In
this way, the capabilities of the model, as received, will be far greater than the executable version
that is finally used for compliance.
A major theme in the industry is that efficiency gains are significant from design integration,
particularly powertrain integration. But it is understood that combined powertrain control is
proprietary. The supporting language might address this more clearly, noting that the GEM
model employs just steady-state maps and a defined set of gear ratios, and cannot predict the
benefits of more sophisticated integration. In a similar fashion GEM cannot predict the benefits
of learning algorithms, look-ahead strategies and intelligent vehicle systems for the optimization
of powertrain efficiency on specific routes. These are emerging approaches, but it is
acknowledged that it would take great effort to configure GEM to deal with these details and it
would be difficult to assure their generic benefit in revenue service. GEM has some check-a-box
options proposed for features that cannot be modeled.
The model, as provided, was oriented to diesel engines. The shift strategies also considered the
engine torque curve for execution. Naturally aspirated gasoline, boosted gasoline and natural gas
engines are likely players in the next five years and may warrant separate and careful
consideration because their characteristics, torque curves and efficiency maps differ substantially
from the diesel engine properties.
b) Performance of each component model including the reviewer's assessment of the
underlying equations and/or physical principles coded into that component;
MATLAB/Simulink remains an excellent basis for the GEM model. It is well suited for the
exploratory and development framework, as well as the production of a more limited executable
model.
GEM may be viewed on three levels. At the highest level, the MATLAB/Simulink platform
allows GEM to be anything it chooses, with the addition or alteration of modules and component
models. At the second level, there is the model provided for review, which has the innate ability
to deal with a wide range of cycles and truck configurations. At the lowest level will be the final
executable version, where certain parameters are fixed, and where the duty cycles are chosen for
determination of compliance. Only the second and lowest levels can be considered in this review.
This GEM model is being produced at a time where engine control strategy development and
integrated powertrain controls are advancing rapidly. Also, transmission options are now far
86
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
wider than in the earlier GEM model, where unsynchronized manual, synchronized manual and
traditional automatic transmissions dominated the marketplace. GEM is challenged in modeling
and giving credit for the technology subtleties that will emerge in the marketplace over the next
five years.
The use of distance, rather than time, as a basis for the test cycles represents a great and
important advance. It was widely recognized two decades ago that light duty automobiles were
capable of far more aggressive performance than was embodied in the FTP-75, although the
FTP-75 was used as the norm, and still embodied allowances for vehicles that could not follow
its modest accelerations. In the heavy-duty arena sustained use of full engine power both on
grades and during acceleration is the norm. All else being equal, a more powerful truck will
complete its duties in less time than an underpowered vehicle, and often the more powerful
choice represents the overall economic optimum. However, the more powerful truck will spend
less time at full power, and will enjoy a reduced average "%load" in revenue service. The
adoption of the distance-based approach is an important step toward matching the GEM model to
real-would use. However, as standards are finalized, the accelerations and grades of cycles
embodied in the GEM execution must represent real life as well. If the grades and acceleration
values are not appropriately challenging, the distance-based approach will look more like a time-
based approach, because underpowered trucks will be able to follow the trace in the minimum
time allowed. This would divorce the market incentives from the environmental incentives and
create a false impression of fuel efficiency capability at the expense of economic reality. It is
important that the grades and test weights used are realistic and representative. Further, the
indication of the ratio (actual / minimum time for cycle completion) is a very beneficial output
value.
A driver's habits are known to have a measurable impact on truck fuel efficiency. This is in part
due to gear selection that influences the engine operating envelope, and in part due to transient
pedal demand (driver PI controller), that may cause an engine to depart from steady-state mode
to a greater or lesser degree. The GEM load demand controller and GEM gear selection
algorithms can be configured to reflect different driving habits. If the chosen GEM driver yields
a better fuel efficiency than would be expected of the national average driver, then GEM will fail
to grant auto-shifting technology and certain engine control strategies their full potential
contribution to efficiency gain. It is recommended that this issue is at least explored to provide
driver sensitivity results to the accompanying document.
The engine model in the new GEM is still a steady-state map. The documentation acknowledges
that transient fuel efficiency will differ from steady-state efficiency at the same speed and load in
time. A transient adjustment factor (TAP) seems likely for the final version, but the use of a
steady-state map with a TAP does not encourage manufacturers to improve transient fuel
efficiency if a generic TAP is assigned. This reviewer appreciates that manufacturers are
unwilling to reveal strategies that represent a substantial fraction of product value, yet they
cannot take advantage of their improvements in the model without this information. Two
solutions are possible:
(i) Determine a manufacturer-specific TAP rather than an agency-specified TAP, or
(ii) Determine an alternate fuel efficiency map for transient (e.g. Heavy-Duty FTP)
operation, and use that map or both steady state and transient maps for the transient cycle.
87
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
In addition, TAP will need to be considered carefully for gasoline and natural gas engines, where
enrichment management has an important effect on fuel efficiency. In a similar way, cylinder
deactivation strategies may be difficult to characterize for throttled engines, because the
deactivation strategy is not well represented by a steady-state map. It is important to have a TAP
strategy that acknowledges successes of manufacturers in lowering transient operation
disadvantages.
The EPA developers have sought comment on the issue of axle efficiency. It can be dangerous to
employ the overall fuel economy computation to compare two approaches to modeling a single
component, particularly if that component represents a small loss. In one case, the overall fuel
economy differed by 1.67% when a simplified efficiency was compared with a lookup table for
rear axle efficiency. This 1.67% represented more than a third of the loss associated with that
component. To place this in further perspective, the 1.67% can be compared to the 3-5%
contributions of a major component, such as side skirts on a semi-trailer at freeway speeds.
Generically, if a component model with higher fidelity is available, in the opinion of this
reviewer, it is worth including the detail in GEM, even if the component model is fixed in GEM.
Furthermore, the existence of such a model may provide a clearer pathway for entering
technology advances by a manufacturer who improves that component in the future.
The tires in GEM are simply characterized by a rolling resistance coefficient and an effective
loaded radius. It is known that the coefficient varies with vehicle speed and contact area, while
longitudinal slip can change the revolutions per mile by a few percent under driving torque and
braking, which could bias results more substantially when grade is present. This could be coded
into GEM, but it is probably more appropriate to treat it as a comment, since detailed tire data of
this kind are not in the public domain and test methods are not universally defined.
c) Input and output structures and how they interact with the model to obtain the expected
result, i.e.,fuel consumption and CO2 over the given driving cycles;
The accompanying explanatory document emphasizes that the input and output structure has not
been finalized. A comment, intended as guidance, is that manufacturers will be obliged to use the
executable version of GEM a large number of times to cover each order or each truck technology
change over the year. This is needed to compute an average value for compliance. The EPA
should make every effort to insure that the final version can be interfaced with the
manufacturers' software to insure that the process is efficient and reasonably inexpensive, while
keeping that version of the model locked to insure compliance.
The .csv output files, viewed in Excel, provide a representation of likely input and output files.
These summaries are very useful and appropriate.
d) Default values used for the input file, as shown in "Vehicle Simulation Model" document.
The default values are not particularly important at this stage of development. They are
sufficiently representative of recent or current technology to provide reasonable inputs and to
assess the ability to predict real-world values. However, this is an appropriate point to
acknowledge and discuss the comparison of GEM output with real-world measurements, as
described in the supporting material.
88
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
The real-world data serve to verify the ability of GEM to model a variety of trucks. These data
do not extend to the validation of truck tire rolling resistance or aerodynamic drag value
selections because the validation was conducted using roller and powertrain dynamometers,
where these values were entered and were the same values used in GEM. Generally GEM
matched the measured fuel efficiency values to within 5%, and the deviation may be attributed
both to measurement error and to the inherent simplifications in modeling. There are some
systematic errors evident in the comparison. For example, the T700 truck efficiency is
underpredicted (GEM vs. measured) on the high speed tests and overpredicted on the low speed
tests. This trend differed for the T270 and F650 trucks. The refuse truck fuel efficiency was
uniformly overpredicted. This leads to the conclusion that GEM may be capable of predicting
overall fuel efficiency accurately (at the 5% difference level), but that one should still be
cautious of comparing the performance of very different technologies using the GEM tool.
The narrative states that "While it is encouraging that GEM accurately simulates overall vehicle
performance in an absolute sense, it is actually more important that GEM is accurate in relative
comparisons." This is true in the sense that GEM should encourage the best technology pathways
through comparison, but it is nevertheless the single overall predicted efficiency that is of
interest to the manufacturer because it predicates compliance. As an example, when the axle
ratios were adjusted for a vocational truck from the chosen value of 3.76 to high (4.06) and low
(3.46) values, the predicted transient fuel economy values were 5.55, 5.49 and 5.6 respectively.
In contrast to this small variation, the fuel economy values for the 65mph operation were 7.14,
6.88 and 7.37 respectively, attributable to the 9% difference in engine speed between the high
and low ratios. This reviewer has confidence in these relative values where a variable input is
changed. However, a comparison between two vehicles with identical chassis and bodies, but
with different engines, transmission types and tire rolling resistance challenges several parts of
GEM in a differential sense, and a greater comparative error must be anticipated. One cannot
argue that the overall agreement with the measured data verifies each sub-model within GEM.
The use of single variables to represent a more complex reality is discussed elsewhere in this
review.
The cycles chosen for evaluating vehicle performance can be changed readily in GEM. Although
incorporation of grades represents a substantial advance, the current choice of fixed 55mph and
65mph steady speeds may cause designers to "teach to the test." [The CBD cycle, used in a
previous age to quantify transit bus performance, suffered from problems of this kind because all
steady state operation was at 20mph.] The output data for the tractor (through engine speed /
vehicle speed ratio) show that the highest gear was used for both fixed speeds. Use of a test cycle
where speeds varied slowly through the 50 to 70 mph range could avoid pitting real world
optima against model optima, and would encourage engine downspeeding strategies that are
successful in revenue service where speed limits are not necessarily 55 and 65mph.
3. When using the standard of good engineering judgment, is the program execution optimized
by the chosen methodologies?
The GEM tool as presented to the reviewers provided a workable compromise between accuracy
and simplicity. It is evident that GEM may be improved in accuracy by increasing input data, in
particular the substitution of data tables for single values. However, these tables would need to
be created for each component at substantial expense. Also, it would be better to use real values
89
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
or tables or strategies from manufacturers, but each of these might require audit and prescribed
measurement methods, and in many cases the manufacturer considers them to be proprietary.
The present GEM is close to being optimized: clearly development is still taking place.
As a general observation, the engine and transmission receive substantial attention in GEM and
in much work on truck efficiency improvement. They seem to be of interest at the single
percentage level. However, tire rolling resistance is a major influence for vocational trucks, and
drag coefficient controls the dominant loss for freeway tractors. Yet these two components are
each represented by a single parameter. Beyond just GEM, a more detailed consideration of these
components (e.g. longitudinal slip of tires, rolling resistance during crosswind correction, effects
of yaw on drag) would assist in raising modeling accuracy. If that cannot be considered
expediently for this version of GEM, it should be considered in the future, or even embedded in
the tables so that it can be applied without altering the code.
4. Please comment on the clarity, completeness and accuracy of the intended output/results
(CO2 emissions or fuel efficiency output file).
The .csv output results are sufficiently comprehensive for the user who is not executing the
program in MATLAB for research and design purposes. Presumably a report similar to the .csv
is anticipated as the executable version output. In fact, to the manufacturer, who is executing this
at time of sale within a larger accounting loop, a very succinct output would be sufficient.
5. In your opinion, are there any procedures or observations that would have added to the
quality of the GEM tool? Any recommendations for specific improvements to the functioning
of the outputs of the model?
Words of caution may help in presenting GEM to the user and to the public. It is important to
state that GEM has acceptable accuracy in predicting the fuel efficiency of a specific vehicle
under specific circumstances. However, it is equally important to state that if GEM is used to
compare two competing, but very different, technology packages, it may not have the fidelity or
granularity to evaluate which is better. GEM may not determine the relative difference between
the two with high fidelity, and that relative difference will depend on the vehicle vocation or test
cycle used.
Although modeling test weights were provided for the example vehicles, there was no
quantitative discussion of choice of test weight for GEM. Test weights receive only brief
mention in the supporting documentation. It is possible that within vehicle classes, or across
regions with different topography, engine size may be selected based on the anticipated load. The
distance based strategy shows an appreciation for this issue, but some compensation with test
weight for engine size could be considered. Perhaps test weights could be selected so that the
time to complete the route remains within reasonable bounds for more lightly powered vehicles,
but at least that highly powered vehicles are acknowledged to be appropriate for some
occupations, loads or regions. The vocational tractor option in Phase I is not a comprehensive
solution.
90
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
III. SPECIFIC OBSERVATIONS
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Page
8
8
10
16
16
17
18
19
22
23
Paragraph
1
2
2
2
2
Table 1-3
2
Fig 1-9 to 1-
13
Table 1-4
End of page
Comment or Question
Axle lookup table is a better approach
"Brakes" Addition of the inertia component to axle is curious/ a
curious description. This is rather a retarding torque. Perhaps the
inertia becomes a force if MATLAB is viewed over time steps. I was
confused by this language.
It would be good to see (as well) the equivalent to Figure 1-3 with
the proposed locked shifting strategies in GEM, rather than actual
strategies as discussed on p. 10.
Need to be cautious about claims in modeling changes. Effect of
change in Crr is clear, bur GEM may have difficulty with relative
accuracy in changing transmission type.
Note in presenting these agreements that Crr or Cd changes are
entered as dynamometer A-B-C coefficients, and these are not real
world measurements. Essentially the dynamometer is partially
modeling these effects.
And the figure. Note that the 1.8% error represents a 17% error based
on the difference, if this figure is intended to show the accuracy of
predicting differences rather than absolutes.
GEM is using automated shifting, essentially, for both MT and AMT,
and this will be used (with PI control pedal) for impact assessment.
Yet this paragraph brings human drivers to the fore. Philosophically,
that the human drivers are the truth and the accuracy of revenue
service, more than the model or powertrain cell.
Are these with default or manufacturer's shift tables?
Little or no shifting occurs in the 55 and 65 cycles. That should be
stated.
Text should explain in a little more detail how these powertrain tests
may be inserted into GEM
Specific Observations on Electronic Model Entitled, "GEM Tool"
Input
Variable
Comment or Question
No Specific Comments or Observations Provided.
91
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Reviewer # 2
92
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Review Comments on EPA's Heavy-Duty Greenhouse Gas Emission
Model (GEM): Phase II and Supporting Documentation
I. GENERAL IMPRESSIONS
This document summarizes the findings of the review of the US EPA's Heavy-Duty Greenhouse
Gas Emission Model (Phase 2 GEM) and supporting documentation ("Vehicle
Simulation Model"). The tool will serve as the principal support for the second round of Heavy-
Duty GHG emissions regulations, which are under development by NHTSA and EPA. The
agencies are considering recognizing the efficiency of various vehicle, engine, and transmission
technologies and they consider critical to develop methods that assess the expected real world
performance of those technologies. The main purpose of this review is to evaluate how well the
developed model can serve as a regulatory and compliance tool. The following represent my
review of the tool and accompanying report based on my experience in modeling heavy-duty
vehicles with full-vehicle simulation tools, as well as from our assessment at the International
Council on Clean Transportation (ICCT) of other heavy-duty vehicle regulatory models used
around the world.
After reviewing the Matlab/Simulink model and the accompanying report, my general
impression is that the "Phase 2 GEM" constitutes a valuable development effort by EPA to
develop a rigorous tool that represents the relative efficiency and emissions of vehicles. The new
modeling tools' comprehensiveness, quality and amount of data inputs, and modeling structure
reflect state-of-the-art modeling techniques and accurately represent relative efficiency
differences of vehicles in real-world conditions.
The model architecture is clear and easy to follow and has incorporated some key features that
will enhance its overall accuracy with respect to real world performance of technologies, and
will allow the model to capture fuel consumption reductions from a broader range of
technologies. Overall, the tool offers a rigorous and comprehensive simulation accounting of
both engine-specific and full-vehicle effects in a manner that is suitable for the regulatory
compliance purposes as indicated. The model will be capable of performing its intended purpose
of reflecting technology benefits for compliance purposes of most of the technologies that the
agencies are considering.
Some new vehicle modeling features are especially important, namely the ability of the model to
incorporate user-defined engine fueling maps and driveline parameters, the development of
different transmission options, the enhanced transmission gear-shifting strategy, the inclusion of
a distance-based routes with road grade, and the more comprehensive treatment of vocational
truck technologies. The accompanying testing effort that was undertaken to validate the model is
impressive and thorough, as capturing the effect of combinations of technologies in such close
agreement with powertrain and chassis dynamometer testing is a difficult task. The model
development demonstrates a thorough development process, and also shows a strong
commitment to transparently presenting the data and methodology that were involved.
The comments below provide additional details, as well as some suggestions that could also be
considered by the agencies in the final model development.
93
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Although the tool itself offers a suitable modeling platform, the document that describes
modeling approach could provide further details in a number of areas. It appears as though the
documentation available for this peer review was at an early draft stage. There is an overall lack
of detail on key technical features that are new in the model. Interested readers would gain from
better descriptions of such features, how they were developed, and perhaps, more quantitative
results in several areas. Also, the quality of the report may be enhanced with more consistent use
of terminology and a reduction in the number of self-references. Further details regarding areas
where such documentation and enhanced information would be helpful are described below.
II. RESPONSE TO CHARGE QUESTIONS
1. Please comment on EPA's overall approach to the stated purpose of the model (meet
agencies' compliance requirements) and whether the particular attributes found in the
resulting model embodies that purpose. Were there critical results or issues that were not
discussed or addressed by the GEM tool or its component sections?
The proposed Phase 2 standards are predicated on the performance of a broader range of
technological improvements than Phase 1, including changes to transmissions and better
integration of engines and transmissions, so a more comprehensive model is required. The model
in its current form will be capable of performing its intended purpose of reflecting technology
benefits for compliance purposes of most of the technologies that the agencies are considering.
The model is enhanced in various aspects with respect to its previous Phase 1 version. Fuel maps
are one of the most important elements in simulation-based models and the new feature of using
actual maps and drivetrain parameters would make the results more realistic, allow the model to
capture the effects of matching engine and driveline, and ideally promote right sizing of the
engines to application. Different transmission options are included in the model. The shifting
behavior is now more realistic since is based on both throttle and speed inputs, and includes the
effects of a clutch friction model. Phase 1 GEM shifting strategy was based only on vehicle
speed and there was no torque interruption during shifting. Road grade has a major impact on
fuel consumption and its addition to the tests cycles would also make the results more realistic.
The treatment of vocational technologies, which were limited to the tires in Phase 1, is
considerably enhanced. The approach followed by EPA to tackle the diversity of vocational truck
applications is appropriate. A few drive cycles are simulated (ARB transient, 55 mph and
65 mph cruise with grade, and a new idle cycle) and weighted differently based on specific
application.
EPA's approach involved a good amount of testing and validation. It must be said that most of
these validation efforts only covered fuel consumption results from reduced weight, better
aerodynamics, and better tires. The validation effort for transmission types and engine-
transmission-vehicle interaction was less comprehensive. However, based on the results
presented, Phase 2 GEM model would be accurate enough to support regulation and drive
technology adoption.
More broadly, I make one final comment on how the overall modeling approach may meet
EPA's overall goals for the regulation, related to the public release of the GEM input and output
data. The existing and Phase 2 heavy-duty vehicle regulation approach relies on the GEM inputs
and outputs to determine compliance. The GEM data are analogous to the light-duty vehicle
94
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
gram/mile1, heavy-duty vehicle gram/brake-horsepower2, and light-duty vehicle mile-per-
gallon3 compliance values. For the heavy-duty use of the GEM in the greenhouse gas emission
regulatory program to meet the agency's own standard, the input and output data from GEM
would ideally be made publicly available just as the regulatory data for the other regulations for
each engine and vehicle. For the public to have confidence in the regulatory program that is built
on a mix of engine- and vehicle-model-specific inputs and modeled GEM outputs, the underlying
data would be presented in full in downloadable data files (e.g., in Excel) as in other EPA
regulations.
References:
1 EPA http://www.epa.gov/otaq/crttst.htm
2 EPA http://www.epa.gov/otaq/certdata.htm
3 EPA http://www.epa.gov/otaq/tcldata.htm, http://www.fueleconomy.gov/feg/download.shtml
2. Please comment on the appropriateness and completeness of the contents of the overall
model structure and its individual systems and their component models (i.e., using the
MATLAB/Simulink version), if applicable, and considering the following:
a) Elements in each system used to describe different vehicle categories;
b) Performance of each component model including the reviewer's assessment of the
underlying equations and/or physical principles coded into that component;
c) Input and output structures and how they interact with the model to obtain the expected
result, i.e.,fuel consumption and CO2 over the given driving cycles;
d) Default values used for the input file, as shown in "Vehicle Simulation Model" document.
Overall the model structure and its systems are appropriate and, in large part, complete.
Generally, the performance of each component model and the underlying equations and physical
principles are valid throughout (see some finer details below). The input and output structures
interact with the model to obtain the expected result in a way that is sound. The following sub-
sections comment on specific issues regarding model structure, individual systems, as well as
default values, in no particular order of importance.
Fixed payloads
Phase 1 GEM had predefined engines, driveline parameters, and payloads for every category. An
issue that may arise when using user-defined engine fueling maps in combination with
predefined payloads is that some simulated vehicles, with lower power-to-weight ratios, will
show higher deviations from the target speed-distance trace. This affects the simulation results
since these underpowered vehicles will take more time to complete the assigned route and will
show a lower average speed. This could lead to underpowered vehicles being improperly
credited.
Appropriate matching of engine, transmission gear ratios, axle ratios, and tire radius is only
going to be promoted if the GEM payloads closely match actual vehicle operation. Right sizing
of powertrains to application does not seem to be promoted when payloads are predefined for a
particular vehicle category. In order to recognize engine power matching to vehicle road load,
payload needs to be a user input rather than a predefined parameter. The regulatory approach and
modeling would ideally recognize and promote market diversity and identify potential
discrepancies between actual payloads and GEM payloads. There is an existing trend towards
95
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
smaller engines, but also some applications require larger engines. On the other hand, if the truck
manufacturer is allowed to input vehicle-specific payloads, some issues may arise in terms of
enforceability (How do the regulatory agencies ensure that the vehicles are operated close to the
payload values at which they were certified?), that may also open the door for the manufacturers
to report numbers for their own benefit, and adds complexity.
An option could be to adjust the payload on a few pre-defined bins based on certain parameters
that are indicative of vehicle road load (e.g. engine displacement, engine power, final drive
ratio). Under this option, a performance criterion that captures the trace-following capabilities of
the simulated truck (e.g. a set threshold of percent difference between target speed and simulated
speed) can be used to force certain engine-vehicle combinations to switch to a lower payload bin
if they don't follow the trace according to the specified criterion. Another option would be to
impose a CO2 penalty based on the ratio of simulated average speed to target average speed.
Ideally, the allowed deviations from the target trace should be minimized for the simulations to
be considered valid and allow comparisons between them.
Drop-down technologies
The agencies have identified a list of technologies that provide fuel consumption benefits but are
difficult to simulate accurately. They are developing feature-based drop-down menus that make
post-simulation adjustments (percent reductions) to the results. It appears that manufacturers
have not taken much advantage of the Phase 1 advanced technology structure to earn credits so it
is important to try to include most of the technologies in some way. However, drop-down menus
inherently assume that all the technology variants within a technology category provide the same
fuel consumption benefits. Not all the models and brands of a certain technology feature would
provide the same fuel consumption benefits. There is the risk of giving artificial credits to
products that perform at a lower level than the value that is selected from the drop- down menu,
thus rewarding poor performers. Also, technology products with better than average levels of
performance would not get additional credits, which is a disincentive to make investments in the
development of such technologies. The default improvement values (percent reductions)
developed by the agencies were not shared for this peer review but they are of relevance and
need to be determined with care. Currently, the users have no flexibility to enter their own
values. Giving the users the flexibility to enter their own values (after testing and with proper
documentation) could offer a way to reward good performers.
It seems that applying adjustment factors in terms of percent reductions rather than applying
predefined credits in units of go2/ton-mile or gal/ton-mile may punish good performers.
Assuming that truck A emits 90 go2/ton-mile and truck B emits 100 gCO2/ton-mile. If a certain
technology improvement value is set at 5%, and both trucks use such technology, truck A would
get 4.5 gCO2/ton-mile credit and truck B would get 5 gCO2/ton-mile credit. This discrepancy of
incentives can exacerbate if the trucks use more than one drop-down technology and the agencies
decide that the percent improvements are additive. So it would be good for the agencies to
support whether and why percentage-based (versus gCO2/ton-mile based) are most appropriate.
Also the agencies might address, in such drop-down menus whether such technology
improvements are indeed additive or not.
Another issue with drop-down technologies is that there is the potential for double counting of
technology benefits. As an example, an electric coolant pump is listed as a drop-down
technology. Depending on the engine mapping process, the resultant engine fuel map may
already capture the benefits from that technology. Running a simulation with such a map, and
96
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
later improving the results using a drop-down menu will double count the benefits. If EPA could
respond to how potential double-counting situations are minimal, that would be helpful.
Driver subsystem
In vehicle simulation modeling, it would seem that the driver ideally would be excluded entirely
as a factor that could influence the GEM regulatory compliance results. Using the same driver
model for all the vehicles seems to be an appropriate choice. However, additional documentation
is needed for this subsystem. There are no details about how the proportional and integral gains
of the PI controller have been selected. Are they representative of current drivers? Are they tuned
to enhance the trace-following capabilities of the model? The look- ahead feature also lacks
documentation. Is it bringing any advantage to the trace-following capabilities of the model?
How was the time span value for such feature selected? Ideally EPA would provide some
consideration and discussion of such factors to provide greater assurance that no anomalies occur
in compliance results from company-to-company technology strategies as well as tested-versus-
real-world results for the relative technology benefits.
Transmission subsystem
There are some transmission-related features that are confusing and need to be clarified. The
report mentions that the different transmission models: manual (MT), automated manual (AMT),
and torque converter automatic (AT) are built of similar components, but each features a unique
control algorithm. However, the model seems to use the same "auto shift algorithm" to determine
the operating gear for any transmission type. The differences in the control algorithm of the three
different transmissions are not clear and need to be provided. Since transmissions are an
important new addition for Phase 2 GEM, it is important to let the reader know that the control
strategy (e.g. shift points) or the selection of predefined transmission parameters (e.g.
efficiencies and inertias at different gears) are not creating any artificial advantage of one
technology type over the others. I suggest presenting a comparison of the same simulated truck
with different transmission types. It is also important to highlight in the report that the new
transmission controller is based on both speed and throttle position, and differs from the Phase
1 transmission controller, which was solely based on vehicle speed. The rule-based approach of
the "auto shift algorithm" would ideally be documented.
It would be appropriate for the agencies to acknowledge that Phase 2 GEM simulations can
capture some but not all of the benefits of powertrain integration. The simulation would
adequately capture engine down speeding since the users have to input specific transmission gear
ratios, final drive ratio, and tire radius. However, there are many complexities in the control
strategy when it comes to integrating engine and transmission. Integrated engine-transmission
powertrain approaches with advanced controls and shifting algorithms that many companies are
developing could result in significantly more (or less) benefit than the agencies determine as the
Appropriate default emission-reduction effect.
As an example, if two different vehicles have the same driveline parameters (tire radius, final
drive ratio, transmission gear ratios, transmission inertias, and transmission efficiencies) and
AMT transmissions from different manufacturers, they will obtain the same simulation results in
GEM but, due to differing control strategies and other design characteristics, they will show
different fuel consumption benefits in reality. It cannot be expected that all the AMT
transmissions bring the same fuel consumption benefits. The drop-down menu option won't
97
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
handle these differences unless there is an option to choose manufacturer-specific transmissions
or otherwise input such data.
As a result, there is an opportunity here to leverage the powertrain testing and provide the option
for manufacturers to better capture the fuel efficiency gains coming from the control strategies
and other complexities that are not adequately captured in GEM. Another advantage of
powertrain testing is that the manufacturers would not need to disclose confidential information.
The results from powertrain testing can then be implemented as correction factors for the GEM
results. Using correction factors, GEM results could be multiplied by a fixed percent
improvement obtained by comparing the results of powertrain test and GEM simulations under
the same torque-speed trace.4 The default benefits for transmission improvements would ideally
be set to be appropriately conservative (i.e., lowest expected value based on various industry
results) in GEM. The drop-down menu could still then be offered as a default, for the
manufacturers that decide not to use the powertrain testing. Then, for the powertrain option,
companies would ideally be provided clear testing procedures and guidance to demonstrate the
emission-reduction impact of their advanced powertrain approaches with physical vehicle testing
in simulated real-world conditions.
Engine fueling maps
The inclusion of manufacturer-specific engine maps is a critical feature to reflect company
differences and detailed engine-specific characteristics that reflect real-world fuel consumption
and emissions. This is an important addition to GEM, but there is lack of documentation of the
engine mapping procedure. I imagine that a fairly prescriptive procedure (including number of
points, preconditioning and warming procedures, fuel properties, etc.) is described somewhere
else in the larger regulatory development document but this chapter would ideally include a brief
description of the procedure so the reader knows which engine accessories are included or
excluded during the engine mapping procedure.
It is noted that there are many advanced features that may affect fueling but are not captured by
using a steady-state fuel map. Manufacturers are going away from traditional map-based
strategies and are going towards model-based controls. Diverse thermal management strategies
are utilized, and some engines use dual torque curves. Have the agencies considered how to
handle these technologies? This could have important implications for how tested steady-state
engine maps, and GEM modeling, and real-world emissions characteristics could differ. As a
result, we recommend that the agencies discuss such industry approaches in the rulemaking and
investigate ways to ensure that tested results are aligned with real-world engine and vehicle
operation the results in fuel consumption and emissions.
The approach used to quantify the transient correction factor (run GEM with the engine map,
then use the torque-speed points in the engine dynamometer and compare measured versus
simulated results) is appropriate. Ideally the transient correction factor may be obtained for each
individual engine. However, since there is a need for selection of a vehicle in GEM in order to
get the torque-speed trace. It would become a hard task for the agencies to try and run a transient
correction factor for each vehicle-engine configuration. For practicality, I recommend
provisionally using a single correction factor and maintaining the option to refine it over the
years with additional testing.
Modeling of idle cycle
98
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
The idle cycle modeling would gain from increased documentation. Using a gCO2/mile value for
an idle cycle at first seems counterintuitive (i.e., there are no miles traveled) so a complete
description of the calculation method would clarify. It would be desirable to present some
validation results for the idle cycle modeled in GEM compared to experimental results. Some of
the engine auxiliaries may not be enabled while doing the test, and the map could be
underestimating actual idle speed fueling rates. There are also engine thermal management
strategies that are used to keep appropriate after treatment system temperatures. These strategies
vary from manufacturer to manufacturer and could increase idle fueling substantially.
The "trace following" issue discussed above also has implications in the calculation of idle cycle
g/mile value. For this calculation the fuel rate in units of grams per hour [g/h] is converted to
units of grams per mile [g/mile] using the weighted average speed over the three non-idle cycles.
The target speed is used for this calculation and not the actual simulated speed, which may
penalize smaller engines. I suggest EPA to consider if this issue might be significant.
Trailers
Although there is a parameter in GEM for trailer tires' rolling resistance, it is not clear how
trailer aerodynamics is going to be modeled in GEM. Trailer aerodynamics can bring about two-
thirds of tractor-trailer aerodynamic benefits, so this is a critical area that requires documentation
and specification of the procedures for the vetting, binning, and including the input data. My
understanding is that the Coda input parameter is for the tractor only (mid-roof and low roof
tractors are tested in its bobtail configuration), or for the tractor using a "reference" 53-ft dry van
trailer (for high-roof tractors coast-down test). Trailer aerodynamic devices can reduce the
overall tractor-trailer combination aerodynamic drag and ideally the Coda used in simulation
should represent the combination. It seems that there is no current provision to include the effect
of trailer aerodynamics as an input in GEM. The report needs to clarify how the GEM model is
handling trailer parameters (including aerodynamics, tires rolling resistance, and weight
reduction) and if the model is going to use a predefined "reference" trailer for all the tractors.
Ideally agencies would give credit to tractor-trailer integrated designs although it would be
difficult for the agencies to ensure in-use compliance of matching of tractors and trailers.
Accessories
There are opportunities for fuel savings from mechanical accessories and electric accessories but
the agencies decided to keep with the Phase 1 approach of having pre-defined and not
customizable power from accessories. If these parameters are assigned default values, there are
no incentives to implement new technologies that could have greater impact. Allowing
accessories power consumption to be user-defined inputs can be used to promote developments
in technologies that reduce the power requirements of accessories such as the alternator, air-
conditioning compressor, power steering pump, or cooling fan. There are other opportunities for
engine accessories such as oil, coolant, and fuel pumps, but is not clear at this point if all those
savings are going to be captured by the engine mapping process.
References:
4 See Sharpe, Delgado, Muncrief (2015) Comparative assessment of heavy-duty vehicle
regulatory design options for U.S. greenhouse gas and efficiency regulation.
http://www.theicct.org/us-phase2-hdv- regulation-design-options
3. When using the standard of good engineering judgment, is the program execution optimized
by the chosen methodologies?
99
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
The chosen methods and execution of the model shows strong engineering judgment throughout.
A good indication of proper execution is the overall good agreement between the Phase 2 GEM
simulations and testing data obtained with chassis and power train dynamometers. The errors
shown are well within +/-5%, which is within the test-to-test variability of chassis dynamometer
testing. Execution at this level of fidelity meets our own criteria that we have utilized to validate
tractor-trailer simulation results.5 the validation results show that the balance between model
accuracy and simplicity is adequate. As a result, the program would be effective to model a
diverse set of technology changes and be used in regulatory applications.
References:
5 See Delgado and Lutsey (2015). Advanced tractor-trailer efficiency technology potential in the
2020-2030 timeframe http://www.theicct.org/us-tractor-trailer-efficiency-technology
4. Please comment on the clarity, completeness and accuracy of the intended output/results
(CO2 emissions or fuel efficiency output file).
The data reports did not appear to be fully complete, but the accuracy of the output/results
appears to meet reasonable expectations. The input and output structure of GEM was not
finalized when released for peer review, however some samples of the output files were provided
to give the reviewer a flavor of the potential structure. In my opinion, for completeness, the
output file needs to include results for each different cycle and not only for the weighted
aggregation of cycles. Some metrics can be added to the output file to facilitate troubleshooting
and give the user a better perspective. As mentioned before, actual simulated speeds and a
measure of deviation from the speed-distance trace would ideally also be provided for
transparency of the results. Based on the validation results, accuracy with respect to measured
data was provided and seems to be within 5%, which is acceptable output accuracy based on
comparable modeling as well as real-world testing.
5. In your opinion, are there any procedures or observations that would have added to the
quality of the GEM tool? Any recommendations for specific improvements to the functioning
of the outputs of the model?
Phase 2 GEM could generate two different output reports. One that only includes the most
relevant information on an aggregated format and is used only for compliance purposes, and a
second one, that is very detailed and includes results disaggregated by cycle and other relevant
information that may help the users to troubleshoot their results, learn the inner workings of the
model and potentially suggests enhancements to it. I suggest including a summary of the energy
audit in the output. Also, provide the average engine efficiency over the cycle for the different
test cycles, as well as the ratio of average engine efficiency to maximum engine efficiency,
which is an indication of how well the transmission parameters are matched to keep the engine
operating near its peak efficiency range.
The output file provides some basic "sanity checks" such as number of shifts, ratio of number of
shifts to number of gears in the transmission, distance traveled, and ratio of actual time to target
time. Please also provide ranges of valid or acceptable values for these parameters so the user
can be aware of any potential issues with the simulation.
100
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
A further step that could allow the tool to be much more useful would be to allow users to input
their own cycles as is currently done with VECTO6 (Vehicle Energy Consumption calculation
Tool) model in Europe. The VECTO tool has a "declaration mode" for compliance, and an
"engineering mode" which offers the ability to edit inputs and allow users to explore what the
tool can do. This would be critical for transparency and follow the best practice as seen in the
Europe situation. It would also be highly useful for individuals in the heavy-duty vehicle supply
chain to explore the variation of the results with respect to real-world duty cycle factors.
Especially considering the very diverse use of heavy-duty vehicles in local, regional, and long-
distance conditions, this capability would allow dealers and fleet managers to gauge how fuel
consumptions for particular relevant driving patterns differs from the cycle. This would help
ensure the technologies that are more suited to particular duty cycles are being selected in the
market place, and it would also help overcome the prevailing market barrier, whereby
knowledge, data, and confidence on truck efficiency has been limited.7
References:
6SeeLuzetal(2014)
http ://ec. europa. eu/clima/policies/transport/vehicles/heavy/docs/fmal_report_co2_hdv_en.pdf
7 See Roeth et al (2015) http://www.theicct.org/hdv-technology-market-barriers-north-america
III. SPECIFIC OBSERVATIONS
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Page Paragraph Comment or Question
The list of key technical features may include the fact that the new
model uses distance-based cycles instead of time-based cycles,
and the fact that test cycles now include road grade.
The claim "more stable engine idle speed controller" is not discussed
in the text. Some metric or quantification of what is meant by "more
stable" needs to be provided.
Regarding "substantial effort has been put forth to accurately track
and audit power flows through the model to ensure conservation of
energy" The report lack details about the energy audit. Was the
energy audit developed just for internal quality control or is it going
to be provided to the end users in an output file?
Regarding "the road gradient has been modified to accept a road
grade that varies as a function of distance traveled" Please
introduce the concept of distance-based versus any pros/cons of
the new method and why did you change the approach.
Driver subsystem. This section (1.2.2.2) is not clear to the reader and,
in my opinion, needs rewriting. There are various confusing
statements such as "the feed forward calculations using drive cycle
accelerations and vehicle mass have been removed". The section also
mentions (page 4, paragraph 1) that a ratio of speeds (which is non-
dimensional) is integrated to produce the current cycle position
(which has units of distance), which is dimensionally incorrect. I
recommend showing the equations to avoid confusing the reader.
101
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Page
O
4
5
5
5
11
5
6
6
8
9
12
Paragraph
2
2
1
O
4
5
1
2
2
2
1
Comment or Question
Regarding the statement "the addition of distance compensation
allows all simulated vehicles to complete an equivalent trip such as
traveling from point A to point B" Does that mean that without
distance compensation the vehicles would not complete the trip?
How were the proportional and integral constants of the PI
controller estimated? Is the same driver subsystem used
independently of transmission choice? How was the look-ahead
feed-forward control implemented?
Consider removing the mention to the "variant power train
architecture" since it is not mentioned anywhere else in the report.
Please clarify that the engine map is not a pre-determined
parameter as in Phase 1 GEM, but a user-defined input.
For consistency with previous section, please show the proposed
constant power loss magnitude of electric subsystem.
Please change "four different variants" for "three different variants"
(MT, AMI, AT).
There is an apparent incongruence in Page 5, paragraph 4, which
mentions that each transmission "features a unique control
algorithm matching behaviors observed during vehicle testing"
however, in Page 5, paragraph 5 it is mentioned that "all of the
transmission models use an auto shift algorithm to determine the
operating gear over the cycle". Are the auto shift algorithm
parameters changed based on transmission type? Is there really a
unique control algorithm for each transmission type?
The "auto shifting optimizer" needs proper documentation. How
does it work?
The clutch model is a key new addition and lacks proper
documentation.
Please support claims such as "realistic actuation durations and
more accurate physics of torque conservation and lockup
behavior" with data.
Please clarify what you mean by "This layout is more similar to a
manual transmission, but the application for a planetary gearbox
is a reasonable approximation as this type of gearbox can utilize a
variety of topologies" is confusing to me.
The statement "The brake model also adds a rotational inertia
component to the axle" is misleading since the inertia of brakes is
set to zero in GEM.
Please provide a table with a complete list of pre-defined and
user-defined parameters. There is no need for specific values, just
the list of parameters.
Description of test condition number 6 reads, "Run a new set of
road load coefficients to represent a vehicle configuration
optimized for fuel efficiency for each vehicle that was tested."
102
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Page Paragraph | Comment or Question
To be consistent with the other test conditions described please
quantify the reductions in rolling resistance, aerodynamic drag, and
mass for this particular "optimized package" case.
13
Figure 1-3 to
1-6
In the legend, it is not clear if the 55mph and 65 mph tests
contain grade or not. Also, the "utility" cycle is not described in
the text.
15
Figure 1-6
Although is mentioned in the text, it seems that the refuse truck
was not tested under the refuse cycle. Also, it seems that it was
not tested under different test conditions as the remaining
vehicles. Any reason for this? Please explain.
16
Change "numerically" for "numerical."
18
The statement "In some cases, it is hard to quantify which method,
vehicle simulation or chassis dynamometer test, is more accurate"
is very misleading. The chassis dynamometer test although
imperfect, is a measurement and it was treated as a "true" value
in the validation process (Figures 1-3, 1-5, and 1-6). Accuracy can
only be measured in reference to a true value. The chassis
dynamometer test was selected as such and the model cannot
hope to have better accuracy than the test.
18
In the statement "GEM is capable of capturing the impact on the
total vehicle CO2 emissions and fuel consumption due to any
technology improvement" the word "any" is misleading since the
validation effort was only done over aerodynamic, rolling
resistance and mass parameters. Moreover, there is certainly a set
of technologies that GEM is incapable of model.
17
Table 1-3
Correct the numbers on the "Delta" column. Due to rounding it is
not evident that the column represents the difference between the
two previous columns. For example in the fourth row we have
3.9% - 4.9% = -0.9%. In the same Table, please calculate the
relative error (consistent with Figure 1-8). Is this relative error
more relevant than the "delta"? What is the maximum acceptable
relative error?
22
Please add a table that summarizes the technologies captured by
the simulation and the technologies that are not captured but are
accounted for via drop-down menus. A third (optional) column
may include technologies that currently are not either simulated,
nor recognized by drop-down menus but potentially may be
included in future regulations.
22
Section
1.4.1.1
Regarding "As described in Chapter 1.2.2, one of the major
changes in the HD Phase 2 version of GEM is to allow
manufacturers to enter their transmission gear number versus gear
ratio" Chapter 1.2.2 does not mention that fact. Please check
throughout the report to avoid issues with these self-references.
103
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model'
Page Paragraph | Comment or Question
23
1
Regarding the statement "Manufacturers also have an option to
select the type of transmission, which is either manual or
automatic" It is not clear if transmission type is a required input
or an optional input. Also, the user should have three options
(AT, MT, and AMT), not just two.
23
Table 1-4
Table heading should mention that the results are from simulations in
GEM, not measurements. Also, it is not clear if the 55mph and
65mph cruise contain grade.
23
The report mentions that due to lack of data, DCT, DCT with TC and
Allison TC-10 AT transmissions were not validated. That assertion
implies that MT, AMT and AT were validated. Since transmission
technologies were not tested at the same level of detail as road load
reduction technologies, it is important to acknowledge the different
level of validation between transmissions and road-load reduction
technologies.
23
Last
Paragraph
Regarding DCT and other transmission types not included in the
model, the report says, "The manufacturers still have the options to
use powertrain dyno tests to quantify the benefits of these or any
other special transmissions". It is not clear if the results of the power
train tests are going to be used to correct the GEM simulation results
or are going to replace the GEM simulation results altogether. Please
clarify.
24
Regarding OEM overriding the axle efficiency input, "the inputs
would be determined by using the prescribed test procedure" It is
not clear which test procedure is the report referring to. Please
add a reference to such procedure.
24
Last
Paragraph
Rolling resistance coefficients are usually expressed in units of
kilograms per metric ton (11 = 1,000 kg). The units used (kg/ton)
may imply short ton (1 ton = 2,000 Ibs.). Please correct and be
consistent throughout the document.
25
The "Regional", "Multi-purpose", and "Urban" composite duty
cycles are not described in the document.
25
Please support your statement, "We concluded that for the 55 mph
and 65 mph duty cycles, GEM's interpolation of steady-state data
tables was sufficiently accurate versus the measured results" What
was the observed accuracy of using steady-state maps for the 55
and 65mph cruise cycles? How much accuracy is sufficient?
25
The 55mph cruise is named "urban highway with road grade", the
65mph cruise is named "rural highway with road grade, and the
ARE transient cycle is named "urban local". Please try to use
consistent names throughout. Also, since the cruise cycles in
Phase 1 were time-based and did not include road grade, the
Phase 2 cruise routes (distance-based with grade) are not
equivalent to them and using the same name is misleading. I think a
distinction can be made by using the term "route" for the
104
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Page
29-30
Paragraph
Table 1-6 &
1-7
Comment or Question
distance-based tests and "cycle" for the time-based, or simply
name them "55mph cruise with grade" and "65mph cruise with
grade".
Automatic transmissions have the same efficiency for all gears
(98%), and such efficiency is equivalent to that of manual and
automated manual transmissions in direct drive. Was this
advantage observed during testing? If possible, please document
with measured data showing the benefits of AT over MT in terms
of gearbox efficiency.
Specific Observations on Electronic Model Entitled, "GEM Tool'
Input
Variable
Comment or Question
File: "GEM_run_postproc.m" Lines: 104 to 108. Potential
issue: It seems to me that the equation is not dimensionally
correct (I might be wrong). Additive terms should have the
same units and the equation seems to be adding gCO2/h
terms with gCO2/mile terms. Please check for missing
terms and/or appropriate use of parenthesis.
File: "GEM_run_postproc.m" Line: 101. The calculation
performed for case 4 is identical to the calculation
performed for case 3, making it redundant. Please
eliminate case 4 (neutral idle with start/stop), or correct the
equation to account for a 90% reduction in neutral idle
emissions (not drive idle emissions).
"vehicle.chassis.frontal
area m2"
The name is misleading. Please rename to "drag area".
Drag area is Coda, aerodynamic drag coefficient is Cd and
frontal area is A.
File: "load_GEM_params.m" Lines: 76 to 85. These
equations set weight reduction penalties for spark-ignited
CNG (525 Ibs.), compression-ignited CNG (900 Ibs.), and
compression-ignited LNG (600 Ibs.). Please set a weight
reduction penalty for the missing case: spark-ignited LNG.
File: "load_GEM_params.m" Line: 94.1 am confused
about the variable "vehicle.chassis.mass_dynamic_kg" as
defined here is a constant. However, during the simulation
(variable: datalog.vehicle.dynamic_mass_kg) the
equivalent mass of the rotational components vary
depending on the active gear.
File: "load_GEM_params.m" Lines: 117 and 143. It is
not clear what exactly the variables
"transmission.autoshift.cost_map" and
transmission. autoshift.required_cost_benefit_ratio"
represent, and how are they used in the model. Other
features such as "restrict skip shifts", restrict shift
105
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Specific Observations on Electronic Model Entitled, "GEM Tool'
Input
Variable
Comment or Question
parity", "disable coast saving downshifts" are not well
documented in the .m files or in the report.
File: "load_GEM_params.m" Line: 131. The
calculation of "Transmission, gear. inertia_kgm2"
involves a multiplication by 0. Therefore it would
produce an array of zeroes. Also, assuming that the
value of 0 is changed to a finite constant, it seems that
the model is assuming the same inertia value for all the
different gears. Is that simplification accurate?
File: "load_GEM_params.m" Line: 134. The calculation
of "transmission.gear.spin_loss_torque_Nm" involves a
division by a factor of 3.73 for all the transmissions that
are not C78_AMT. I am wondering what that factor
represents. Also, if that factor is still valid for a potential
C78 MT transmission.
Energy Audit. Net system kinetic energy change is 0 kJ,
which is the result of the test cycle starting and ending at 0
mph speed. The test cycle seems to start and end at the
same altitude (symmetric grade profile traveled at constant
speed), so one can expect the net system potential energy
change to be OkJ as well. Since both kinetic and potential
energy are conservative and not dissipative forms of
energy, I was confused about the energy audit accounting
for energy "consumed" by gradient at about 17% of the
losses for a tractor-trailer. Checking the equations, it
seems that the energy audit is performed only for positive
tractive loads. If that is the case I am confused about the
energy consumed by the brakes at about 7% of the losses
for a tractor-trailer. They should be really low if one is
only considering positive tractive loads. The brakes are
mostly applied for negative tractive loads (e.g. driving
downhill). Am I missing something here?
Why the weighted average speed
(weighted_avg_speed_mph) used in post processing (file:
"GEM_run_postproc.m") is based on the target speed-
distance trace and not in the actual simulated speed-
distance trace? This decision has implications for
underpowered vehicles that deviate substantially from the
target trace since their actual average speed may differ
substantially from the target speed. This also has
implications for the calculation of idle fuel rates for
vocational vehicles since the conversion factor from idle
fuel rate in [kg/h] to [kg/mile] is the weighted average
speed.
106
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
107
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Reviewer # 3
108
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Review Comments on EPA's Heavy-Duty Greenhouse Gas Emission
Model (GEM): Phase II and Supporting Documentation
I. GENERAL IMPRESSIONS
The EPA Heavy-Duty Greenhouse Gas Emission Model (GEM) Phase 2 documentation
accurately represents the structure, format, logic and algorithmic description of the model as
presented. The supporting documentation is for the most part, clear and self-explanatory. The
results produced by the GEM model appear to be sound, although each set of results is presented
as integrated fuel efficiency and carbon dioxide emissions results. As such the results present a
single-valued, integrated snap-shot of the model prediction, and the time-based instantaneous
fuel efficiency and carbon dioxide emissions predictions are not available for review. The overall
dynamic performance of the model prediction is thus difficult to judge in the greater context of
what is usually fully transient vehicle operation. Furthermore, in the version reviewed, the ability
to vary input parameters and vehicle and drivetrain attributes is limited to modifying input files
and not through a graphical user interface as described in the review instructions.
II. RESPONSE TO CHARGE QUESTIONS
1. Please comment on EPA's overall approach to the stated purpose of the model (meet
agencies' compliance requirements) and whether the particular attributes found in the
resulting model embodies that purpose. Were there critical results or issues that were not
discussed or addressed by the GEM tool or its component sections?
The model appears to meet the stated purpose for which it was intended, which is the prediction
of integrated cycle-based vehicle fuel efficiency and carbon dioxide emissions for vehicles with
preselected physical and drivetrain attributes. While this was the subject of some detailed
explanation in the model documentation, the assumption of quasi-steady engine fueling and its
extension to fully transient engine operation is not without complexity in the assessment of its
validity. A further, acknowledged inadequacy of the model in its current form is the limited
ability of the user to modify specific vehicle attributes, and component values and efficiencies.
2. Please comment on the appropriateness and completeness of the contents of the overall
model structure and its individual systems and their component models (i.e., using the
MATLAB/Simulink version), if applicable, and considering the following:
a) Elements in each system used to describe different vehicle categories;
The elements in each of the systems (engine, transmission, axle, vehicle attributes etc.) seem
appropriate and complete. The specific selection of the engines and transmissions chosen will
cover a large portion of the current heavy-duty vehicle fleet, although of course any specific,
single selection of powertrain hardware or powertrain hardware attributes necessarily limits the
range of vehicles that can be simulated with that same selection.
b) Performance of each component model including the reviewer's assessment of the
underlying equations and/or physical principles coded into that component;
109
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
One concern with the structure and form of the model is that a steady-state engine fueling map is
used in each case to simulate transient engine operation and hence dynamic vehicle operation.
This is in addition to the simulation of operation under nominally steady vehicle speeds or cruise
operation. In the case of nominally steady operation, the use of a steady-state fueling map is
well-justified, but the quasi-steady assumption required to allow the extension of the use of such
a map to transient operation requires additional justification. Heavy duty compression ignition
engines have high rotational inertias (due to the relatively high mass components required to
survive the high combustion pressures), high mechanical friction (due to high effective
compression ratios) and relatively slow air and exhaust transfer processes (due to the excess air
flowrates accompanying their lean, un-throttled operation). In addition they have relatively high
thermal mass due to their large physical mass required to withstand the stresses and strains
resulting from high combustion pressures. All of these features conspire to result in deleterious
combustion effects under highly transient engine operation. In most cases the end effect of these
phenomena is to reduce the engine brake thermal efficiency under transient modes, beyond that
which would be expected under steady or quasi-steady fueling operation. In most cases the
additional fuel that is required to undertake a specific engine transient torque trajectory, beyond
that estimated using a quasi-steady fueling assumption, would typically be less than 10% of the
total integrated fueling, but in most cases the effect on integrated fuel efficiency over transient
duty cycles is non-negligible.
In general, incorporating an additional component into the full accounting of the vehicle load by
over-accounting for the actual total rotational inertia (in the form of an effective or "added" mass
or inertia) of the drivetrain and driveline system does allow for the quasi-steady assumption to
hold. However in general a quasi-steady, forward-looking simulation such as is used in this
model, tends to under-predict the actual vehicle energy usage under transient duty cycle
operation. I notice that this issue is addressed in Chapter 1.4.1.8 Transient Adjustment Factor,
but the designation of a single correction factor for a specific engine or powertrain configuration
is likely to be unsuitable in some cases, and has the potential to cause prediction inconsistencies.
Note further that the required correction factor might not be uniquely engine-specific, but might
vary for the same engine in different vehicle and powertrain configurations.
c) Input and output structures and how they interact with the model to obtain the expected
result, i.e.,fuel consumption and CO2 over the given driving cycles;
A further concern that is not discussed in the documentation is whether any model fitting
parameters were employed to obtain the fits observed, between the GEM-derived fuel efficiency
and CO2 emissions values and the dynamometer-measured results. In other words, beyond the
parameters described and the accompanying constants used in the dynamic force and energy
equations, were any other fitting techniques (or fitting parameters) used to obtain the observed
correlations between the simulated cycle-averaged results and the SwRi chassis dynamometer
results? Presumably there were dynamometer parameters and coefficients fitted using vehicle
coast-down data (and dynamometer operational parameters), but beyond these, are there any
other fitting parameters used to obtain the correlations shown? Discrepancies between simulated
and measured results of less than 2-3% are probably not significant, except in the presence of a
consistent bias between the measured and predicted, for any one vehicle, cycle or technology
considered.
d) Default values used for the input file, as shown in "Vehicle Simulation Model" document.
110
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
The default values as defined in the "Vehicle Simulation Model" are reasonable and currently
fall within the ranges of expected vehicle values. It is not clear however under what
circumstances the user will be able or allowed to make modifications in the final model
implementation. For instance, it is conceivable that the interplay between future auxiliary
mechanical or electrical loads on an engine might be significantly modified through conversion
of mechanical auxiliaries to electrical or electronic devices. In that case, that will shift the
relative balance in those loads. Moreover it is not clear that engine cooling fan loads have been
adequately accounted for in the model, as these are typically not considered in the engine
dynamometer testing from which engine fueling maps are normally derived. This exclusion
alone can modify observed fueling rates by 10% or more under specific engine operating
conditions.
Other default values including transmission gear ratios, transmission efficiencies, axle
efficiencies, tire rolling resistance, and vehicle aerodynamic drag product seem reasonable.
3. When using the standard of good engineering judgment, is the program execution optimized
by the chosen methodologies?
This issue is difficult to address through the level of observation afforded to the reviewer at this
stage of development of the model. It is not obvious that the program execution is optimized, but
the results, computational time and outputs displayed indicate that the chosen methodologies are
suitable for this purpose.
4. Please comment on the clarity, completeness and accuracy of the intended output/results
(CO2 emissions or fuel efficiency output file).
The model outputs and results seem clear, complete and accurate. One caveat with interpreting or
using the freight efficiency or load efficiency-based results lies in the use, further interpretation
or extension of these results. Users might be tempted to "scale" the load-based results in an
inappropriate fashion - for example, if the returned result for the computed carbon dioxide
emissions is 100 gCO2ptm (grams CO2 per ton-mile) for say a 30 ton vehicle over a specific
cycle, there might be the temptation on the part of users to employ that same numerical result to
predict the CO2 emissions for the same vehicle loaded to 40 tons over the same cycle. However
this assumption is not correct, as the vehicle fuel consumption is a function not just of load or
weight-related terms (rolling resistance, grade, acceleration etc.) but also terms that are invariant
with load or weight (such as aerodynamic drag), and this is not reflected in an emissions per ton-
mile result.
5. In your opinion, are there any procedures or observations that would have added to the
quality of the GEM tool? Any recommendations for specific improvements to the functioning
of the outputs of the model?
The version reviewed here does not include the graphical user interface (with "pull-down
menus") described in the instructions. The ability to modify input parameters and vehicle
attributes will improve the user experience, while obviously presumably not impacting the model
outputs.
Ill
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Due care and attention should be paid to the number of significant digits presented in the output
results. For example, in the results presented from a single specific simulation (shown below):
» GEM_Phase2_Idle_5 5_65_CARB_HHDDT_Transient
Distance = 32.414 mi
Fuel Consumption = 5.4059 gallons
Fuel Consumption = 17184.1217 grams
Fuel Economy =5.996mpg
Fuel Consumption = 530.140 g/mile
CO2 Emission = 1697.77 g/mile
The number of significant digits in the model simulation outputs presented above (some of which
are directly related through derivation or calculation) varies from 4 to 9. This does not meet
recommended practices in the presentation of results and data. Moreover, industry experience in
the measurement of real-world fuel efficiency during over the road truck testing dictates that
measured fuel consumption variations of less than 1-2% should not be considered significant or
compelling, and this level of variation could correspond to variations in the 2nd or 3rd significant
figures in fuel consumption in most cases.
III. SPECIFIC OBSERVATIONS
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Page
1
Et seq.
5
5
7
7
8
9
9
10
11
Paragraph
1.2.2.3.1
1.2.2.3.1
1.2.2.3.3.7
1.2.2.3.4
1.2.2.4
1.2.2.4
1.2.3.2
1.3
Comment or Question
Use "dynamometer" and not "dyno"
Use "Phase 2" and not "Phase II"
Use "watts" or "W" and not "Watts"
"it may o make use" requires clarification
"With the new gear engaged the clutch is reengaged and the engine is
again allowed to operate at full load." This statement presupposes that
the transmission was shifting under full engine load, which is not
necessarily the case for high power engines under benign operating
cycles.
"This includes drive shafts as well as driven and passive axles,
consisting of a differential, brakes and tires." Passive axles will not
ordinarily include a differential, only driven axles will.
"The vehicle system consists of the chassis, its mass and forces
associated with aerodynamic drag and changes in road grade". Why
"changes in road grade"? Any constant road grade (other than zero) will
have an effect on the apparent vehicle load, and not just changes in
grade.
"computes acceleration [not accelerations] from the input force and
equivalent mass which is integrated to generate vehicle speed and
distance traveled"
Use "Matlab" or "MATLAB" and not both.
"Validations use all actual vehicle variables conducted at Chassis dyno
cell," needs editing.
Ref. 3 is not the SwRI report as stated. It is an ASME Technical
Paper.
112
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Page
12
13
20
21
22
22
23
Paragraph
1.3.2
Fig 1-11
Fig 1-13
Comment or Question
A +-15% variation in aerodynamic drag, for example, is unlikely to
span the full range of expected values in the future for HD vehicles.
Presumably aerodynamic modifications to Class 8 vehicles may
result in significantly lower drag coefficients, or drag products (CdA)
For the Class 8 T700 tests, at high fuel efficiency, the GEM model
appears to consistently under-predict the actual measured vehicle fuel
economy. This consistent offset is of concern. The reverse is
observed for the Class 6 truck tests.
Caption is incorrect.
Vertical axis incorrect unit designation.
Do not use the term "aero drag". It should be "aerodynamic drag".
"tire radius" should be "tire rolling radius" or 'effective radius'.
"Lack of testing data for other types of transmission, GEM would not be
able to be validated in time against those three cases" requires editing.
Specific Observations on Electronic Model Entitled, "GEM Tool"
Input
Variable
Aerodynamic
drag
Engine
Accessories
Comment or Question
- Audit data for
GEM_Phase2_Idle_5 5_65_C ARB_HHDDT_Transient drive
cycle —
"Energy Consumed by Cd" should refer to "CdA" and not "Cd"
alone.
Engine Accessories = 846.16kJ 0.23%
It seems as though engine accessory loads are under-accounted
for in this implementation of the model - a 0.23% loss for a full
transient cycle seems inappropriately low.
Usable System Energy Provided = 373988.49 kJ
Engine Energy = 309402.35 kJ
Engine Efficiency = 42.08%
This integrated "engine efficiency" is high for the average
efficiency expected across a fully transient cycle - does this refer
to the peak engine efficiency encountered?
113
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Reviewer # 4
114
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
Peer Review Comments on EPA's Heavy-Duty Greenhouse Gas Emission
Model (GEM): Phase II and Supporting Documentation
I. GENERAL IMPRESSIONS
Accuracy of information:
The document provided, "Vehicle Simulation Model" provides a good background on GEM-II,
its differences from GEM-I, and the Phase I certification process. Section 1.2, "Model Code
Description" describes the model components and sub-components, in adequate detail for the
user to understand the depth and breadth of GEM-II. No underlying equations are provided. The
section on Model validation is an important section for GEM-II. The extent of the validation and
the comparisons with dynamometer testing is impressive. This gives the reader additional
confidence in the results produced by GEM-II. The validation process is well documented,
concluding with the graph that summarizes all the 130 vehicle validations performed. The
validation also includes graphs of component performance (engine speed, engine fuel rate, and
transmission gear number) as a function of time
The document does not explain the variable target.veh_sytyle. The structure format used for the
user input data is useful in collecting all user-provided data.
A final conclusions section is missing in the document provided.
II. RESPONSE TO CHARGE QUESTIONS
1. Please comment on EPA's overall approach to the stated purpose of the model (meet
agencies' compliance requirements) and whether the particular attributes found in the
resulting model embodies that purpose. Were there critical results or issues that were not
discussed or addressed by the GEM tool or its component sections?
EPA's overall approach to meet the agency's compliance requirements consists of making a
validated simulation model (GEM -II) available to OEMs so that they can check the compliance
of their vehicles against the agency's guidelines. The extensive validation of GEM - II against
dynamometer testing of the actual vehicle shows a good correlation between simulation and
hardware to within ±5%. Most validations are well within ±3%. This provides the user with a
high level of confidence that the physics has been correctly implemented and there are no
unresolved "bugs" in GEM-II.
Further, an additional level of confidence is achieved, with select validation from four
representative vehicle classes, namely: Class 8 Kenworth T700 truck, the Class 6 Ford F650 tow
truck, the Class 6 box truck, and the New Flyer Refuse truck. There is a very good correlation
between GEM-II predicted engine speeds and transmission gear shifting versus the same on the
actual vehicle.
Recommendation: It is recommended that representative vehicles from each of the classes be
modeled in GEM-II and validation results presented similar to 1.3.1 of the GEM-II Manual. It is
the understanding of the reviewer, based on the provided manual that the following trucks were
115
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
tested on a vehicle chassis: Class 6 Kenworth T270, Class 6 Ford F650, Class 8 Kenworth T700,
Class 8 Cascadia Line Haul truck, and Class 8 New Flyer refuse truck.
Issues/Results that were not addressed by GEM-II:
During this review, the following unaddressed issues were identified for GEM-II:
• GEM-II includes several PID controllers within its overall structure. For example, the
engine idle speed controller of the engine is implemented as a PID with three gains. The
unaddressed issue in this regard is stability feedback to the user caused by unrealistic
hunting for the idle speed. What safe-guards are in place within GEM to inform the user
of clutch chattering since GEM does not model second order inertial effects caused
during clutch engagement.
• It is possible for the vehicle not to meet the driving cycle as a result of excessive grade or
weight or other issues with the transmission/engine. The unaddressed issue in this regard
is a feedback alert to the user during the time instances when the vehicle is significantly
slowed down and does not meet the desired driving cycle. The output file does not alert
the user on the number of time instances when vehicle tracking was compromised.
• GEM-II does not address thermal characteristics of the engine cooling system or the heat
rejection of the transmission fluids. These thermal issues affect the operational duty-cycle
of the engine fan, which will affect the fuel economy. At present, GEM-II models the
parasitic loads as a constant average number.
Recommendation: Allow the user the ability to introduce an engine load dependent
mechanical accessory curve which is more realistic than a constant average number. A
simple heat model may be used to capture the effect of thermal characteristics of the
multiple radiators in a typical MD and HD engine.
• Although GEM-II does not model tire slip/lockup during a hard deceleration, the effect of
ignoring this on fuel economy is negligible.
• The validation results included kinematic comparisons (speeds, gear number) between
GEM-II and the actual vehicle on a chassis dynamometer. While, the kinematic
comparisons look very favorable, the dynamic comparisons (engine load and engine
fueling) are missing in the results section of GEM-II.
• The transmission shift strategy can affect fuel economy and emissions. GEM-II allows
the user to preselect different transmission types (manual, automatic or automated-
manual). However, it was not clear how to modify the GEM-II default shift strategy with
an OEM proprietary shift strategy.
• Although GEM-II results have been extensively validated against dynamometer test data
in a controlled lab environment, it is unclear how well GEM-II will compare against real
world road testing, especially with temperature fluctuations. For example, the lack of a
thermal model in the engine model may cause GEM-II results to deviate from on-the-
road test data, where the engine fan is cycled on and off based on thermal loads on the
engine. Each time, the engine fan turns on, fuel economy is affected.
2. Please comment on the appropriateness and completeness of the contents of the overall
model structure and its individual systems and their component models (i.e., using the
MATLAB/Simulink version), if applicable, and considering the following:
116
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
The three main powertrain components that can affect fuel economy and greenhouse gas
emissions in a vehicle are: engine fuel map, transmission type and efficiency map, and vehicle
aerodynamic improvements (including tire rolling friction improvements and weight reduction
technologies). In this regard, GEM-II addresses all the aforementioned components by providing
steady state maps for each powertrain component, which an informed user can change to
represent specific technology improvements. GEM-II comes with certain standard transmission
models, namely: manual, automatic, and automated manual transmissions. The user would select
the appropriate transmission and GEM-II would automatically select the user-specified
transmission.
Recommendation: It is recommended that additional instructions are provided if the user wants
to change the engine map. Ideally, this would be done from a user specified spreadsheet in a
GEM-II compatible format, since the user may not be fluent in Matlab. In this regard, a clear
explanation of all the variables used in GEM-II would also help significantly. For example, it
was not clear how to change the transmission shift schedule if an OEM chose to do so. Further,
since GEM-II is modular with hierarchical layout of component layers, it is challenging for an
OEM user to insert a technology improvement deep within one layer and not affect the layers
above or the execution of GEM-II. It is not clear, from this initial review, how a technology
improvement such as a proprietary transmission shift schedule can be evaluated in terms of gains
in fuel economy and greenhouse gases. This comment applies to other technology improvements
as well, such as partial engine cylinder deactivation (power on demand) or electrification of
certain mechanical accessories.
a) Elements in each system used to describe different vehicle categories;
The vehicle categories that GEM-II addresses range from Class 2B to Class 8 HD conventional
vehicle powertrains. This is achieved by four root-level systems in GEM-II at the root level,
namely: the ambient, driver, vehicle, and powertrain modules. Each of the aforementioned
modules consists of several sub-modules organized in a hierarchical manner. Each root-level
module outputs a data bus that is muxed into a single data bus. The aforementioned four main
systems of GEM-II correspond to the four main components of a HD vehicle, namely: driver,
ambient conditions, vehicle chassis and powertrain modules. This one-to-one correspondence
between the root-level GEM-II models and an actual HD vehicle makes for an easily
understandable structure. Further, the modular organization of the GEM-II contributes to easier
debugging and isolation of a numerical problem during simulation.
The powertrain module is the most populated module in GEM-II. It contains the engine,
transmission and driveline sub-modules and accessories (mechanical and electrical). The flow of
data information corresponds to an actual vehicle powertrain, with the engine output driving the
transmission, which in turn drives the driveline components.
The modular layout of GEM-II, its correspondence with a real conventional vehicle is therefore
appropriate and complete for the reasons stated above. The modular structure and the
hierarchical arrangement of modules to mimic a real vehicle system makes the integration of
additional modules and capabilities easier to implement. The signals are clearly marked and
follow a logical naming convention that facilitates the addition of additional modules and
capabilities into GEM-II.
117
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
b) Performance of each component model including the reviewer's assessment of the
underlying equations and/or physical principles coded into that component;
GEM-II follows the model of a single wheel with a concentrated mass at the center of the wheel.
The physics coded into the modules are based on Newton's second of motion for this
concentrated mass. At the time of this review, no other dynamic equations were found in GEM-
II, except in the clutch and torque converter. The engine, accessory, transmission and driveline
components are characterized by steady state maps. The equivalent mass of rotating inertia
components are also correctly included in the vehicle module of GEM-II. Rotating inertias are
correctly reflected downstream to the tire. The inertia is converted to a virtual mass which is
added to the entire vehicle mass. The wide validation of GEM-II against real vehicle data
indicates that the physics has been correctly implemented in GEM-II.
c) Input and output structures and how they interact with the model to obtain the expected
result, i.e.,fuel consumption and CO2 over the given driving cycles;
Input for GEM-II:
A structure format is used to store the inputs to create the input data for the execution of GEM-II.
The structure format is organized as follows: component.variable.units. For example, the input
variable "engine.idle_fuel_map_speed_radps" identifies the engine speed vector of the engine
idle fuel map expressed in rad per sec. Similarly, the input variable
"transmission, clutch.input_inertia_kgm2 refers to the clutch inertia of the transmission,
expressed in kgm2. This format follows good coding standards, making the inputs easy to pair to
the appropriate component it refers to, the particular variable name, and the units used. Further, a
modular approach is used to store the input data for each component in separate easily identified
files in the "param_files" folder.
Output for GEM-II:
When the workspace has been populated with the input data, the simulation model "REVS_VM
vehicle model is executed over the user-selected drive cycle. Each major component model
(GEM_CVM, vehicle, driver, ambient) has a bus_out output port which contains a structure of
component output data, which is used within other components. In addition, GEM-II uses a
datalog structure to store simulation output data for later post processing to calculate emissions
and fuel economy. All the simulation output is stored in a single datalog structure with multiple
fields, each describing the component that the data pertains to. For example,
datalog.vehicle.speed_mps refers to data log from the vehicle component of the variable vehicle
speed in m/s. This format is an accepted coding standard within other vehicle simulation
packages (such as PSAT from ANL, RAPTOR from SwRI) as well, making the output easy to
pair to the appropriate component it refers to, the particular variable name, and the units used.
Similarly, all variables that are used to perform an energy balance are prefixed by "audit".
Interaction with the model to obtain the expected results:
The bus_out structure from the various components are stored in "goto" blocks which are paired
with "from" blocks to distribute data from one block to another. The use of paired "goto" and
"from" blocks is an accepted method of decluttering the simulation model and avoiding
crisscrossing signal lines, thereby significantly facilitating the understanding of information flow
from one component to another.
118
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
d) Default values used for the input file, as shown in "Vehicle Simulation Model" document.
The default data used in GEM-II is complete and appropriate to execute a simulation of HD
vehicle powertrain over one of the drive cycles available in the GEM-II default drive cycle
library. The default ambient conditions summarized in ambient_param.m are appropriate. The
default driver parameters, summarized in driver_param.m, contain driver gains as well as the
time that the driver can look ahead in the drive cycle. The driver gains represent an average
driver, which the user can change to emulate an aggressive driving pattern versus a calmer
driver.
The default engine maps of 270 kW, 345 kW, and 455 kW power ratings includes inertia, idle
speed. Default transmission maps for the manual, automatic, auto-manual are also available.
Default tire radius, axle ratios, rolling resistance of the tires of the steering axles and drive axles
are also included.
3. When using the standard of good engineering judgment, is the program execution optimized
by the chosen methodologies?
Overall, GEM-II uses industry accepted coding practices throughout the software modules. The
following is a partial list of these accepted practices:
• Valid variable naming structure used -
• Data bus used for each component -
• Modular components with no signals crossing -
• Useful comments to assist the user with following the code -
• Energy audit adds to the confidence level of the results of the simulation.
• File name that is executed is echoed back to the user. If there is a simulation abort, the
debug is easier
• Data that is being loaded is echoed back to the user so the user knows what data is being
used.
• Component modules are linked to libraries. A change in the library module propagates to
all vehicle models during execution.
At this point in the review, I do not have adequate data to comment on the execution
optimization of GEM-II. Linear interpolation modules from the standard Simulink library are
used within GEM-II, thereby optimizing execution. If there are any non-standard user defined
functions used within the GEM-II simulation model, the execution can be made considerably
faster through the use of s-functions within Simulink. At the time of this review, no S-function
were found in the model.
4. Please comment on the clarity, completeness and accuracy of the intended output/results
(CO2 emissions or fuel efficiency output file).
The output data from a simulation execution is summarized in a spreadsheet, which is date and
time stamped, allowing the user to verify that the output data corresponds to the simulation
executed. The output data contains data on which technology improvement (weight reduction,
vehicle speed limiter, single drive axle, par time single drive axle, low friction axle lubrication,
predictive cruise control, high efficiency AC compressor, electrified engine cooling pump,
119
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
extended engine idle reduction, automatic tire inflation )was assessed, engine, transmission type,
fuel economy and CO2 emissions.
The output data, summarized above, serves the original purpose of GEM-II, which is to enable
users to demonstrate compliance with regulatory standards either without any modifications to
the HD vehicle, or analyze the fuel economy and CO2 emissions, when one or more technology
improvements are employed. However, information on the drive cycle is missing or not clearly
identified. In addition, key plots of vehicle tracking the drive cycle, engine operating speed-
torque points over the drive cycle, engine efficiency contour plots, transmission operating points,
and other plots that assist OEMS to further fine tune the powertrain and improve fuel economy/
CO2 emissions in case of non-compliance. It would be desirable to have the results of the energy
audit summarized in the output file. The date stamp column is appropriate for the user to cross
check simulation runs.
5. In your opinion, are there any procedures or observations that would have added to the
quality of the GEM tool? Any recommendations for specific improvements to the functioning
of the outputs of the model?
The following modules will enhance GEM-II:
• A module that is able to create the input data for a GEM-II execution from a user
provided spreadsheet with pre-defined tabs for the engine, transmission, drive cycle,
vehicle parameters, and technology improvement. Users are more familiar with
spreadsheets than the Matlab environment.
• A GUI module that guides the user to create a HD vehicle model.
• A module that allows users to select plots of key component performance. These plots
may be summarized in the output spreadsheet data file.
• A more detailed explanation of all the user provided data "target.X" and the various
choices available for each of the user provided data. For example, what are the choices
that the user has for the variable target.veh_style ?
• GEM-II execution takes place within the Simulink environment. During execution, no
feedback is provided to the user on the status of the simulation. The user is waiting on a
blank screen - Percent complete of the simulation and which drive cycle is being
executed would be useful feedback for the user.
• Predefined sample input data for all class of vehicles to assist users to easily modify them
if necessary, since some users may not be familiar with Matlab.
• The ability to turn on the feed forward term for the driver model in case of tracking
problems in drive cycles with grade.
• The ability to model accessory power draw as a function of engine speed and engine
temperature. The engine cooling fan power cycle will affect fuel economy and CO2
emissions.
120
-------
External Peer Review of EPA's Heavy-Duty Greenhouse Gas Emission Model (GEM): Phase 11 and Supporting Documentation
III. SPECIFIC OBSERVATIONS
Specific Observations on Tool Description Entitled, "Vehicle Simulation Model"
Page
5
3
6
9
Paragraph
2
2
4
3
Comment or Question
The last sentence reads "If a manufacturer uses a hybrid powertrain
for the power take-off devices, it may o make use of ....". The "o"
after may is a typo
Distance compensation is critical for all vehicle simulations -
Therefore, this is a good feature that has been implemented in GEM-
II
Please explain what is included in spin losses, since this may not be
clear to all OEM users.
The GEM-II executable is very appropriate for users who are not
fluent in the Matlab/Simulink environment. Further, the executable
prevents users from making any changes to GEM-II to support
compliance.
Specific Observations on Electronic Model Entitled, "GEM Tool"
Input
NA
Variable
NA
Comment or Question
Avoid taking the derivative in the Simulink models. This can cause
instabilities if the signal fluctuates rapidly.
121
------- |