Title: Effect of Mass on Multimodal Fuel Consumption in Moving People and Freight in the U.S.

Authors: John L. Sullivana, Geoffrey M. Lewisb *, and Gregory A. Keoleianc

a, b, c
Center for Sustainable Systems
School for Environment and Sustainability
University of Michigan
Dana Building
440 Church St.
Ann Arbor MI, 48109-1041
a
jlsulliv@umich.edu, b*glewis@umich.edu, cgregak@umich.edu
corresponding author
734-936-2637
Abstract

The United States transportation sector consumes 7 billion barrels of petroleum annually to move
people and freight around the country by car, truck, train, ship and aircraft, emitting significant
greenhouse gases in the process. Making the transportation system more sustainable by reducing these
emissions and increasing the efficiency of this multimodal system can be achieved through several
vehicle-centric strategies. We focus here on one of these strategies – reducing vehicle mass – and on
collecting and developing a set of physics-based expressions to describe the effect of vehicle mass
reduction on fuel consumption across transportation modes in the U.S. These expressions allow analysts
to estimate fuel savings resulting from vehicle mass reductions (termed fuel reduction value, FRV),
across modes, without resorting to specialized software or extensive modeling efforts, and to evaluate
greenhouse gas emission and cost implications of these fuel savings. We describe how FRV differs from
fuel intensity (FI) and how to properly use both of these metrics, and we provide a method to adjust FI
based on mass changes and FRV. Based on this work, we estimate that a 10% vehicle mass reduction
(assuming constant payload mass) results in a 2% improvement in fuel consumption for trains and light,
medium, and heavy trucks, 4% for buses, and 7% for aircraft. When a 10% vehicle mass reduction is
offset by an increase in an equivalent mass of payload, fuel intensity (fuel used per unit mass of payload)
increases from 6% to 23%, with the largest increase being for aircraft.

Keywords: transportation energy; transportation sustainability; fuel consumption; fuel

intensity; fuel reduction value

© 2018. This manuscript version is made available under the Elsevier user license
http://www.elsevier.com/open-access/userlicense/1.0/
 1. Introduction

The transportation sector of the United States uses 70% of U. S. petroleum consumption

(about 7 billion barrels per year), which amounts to 28% of total U.S. annual energy

consumption (ORNL 2009b). This energy enables the movement of both people and freight.

However, the combustion of petroleum generates fossil carbon dioxide (CO2) and other

greenhouse gas (GHG) emissions. Reducing GHG emissions has become an imperative that has

resulted in major regulatory policies, such as automobile GHG emission standards developed by

the Environmental Protection Agency-EPA (USEPA and USDOT 2012). Emission reductions are

needed from all modes of transportation including cars, light duty trucks, medium and heavy

duty trucks, railroads, buses, aircraft, and ships. There are several vehicle-centric ways to

achieve reduced fuel consumption (FC, in units of gal/100 mi) and resultant carbon emissions

from the transportation sector, including more efficient conventional vehicle powertrains,

advanced powertrains, and reduced-mass (or lightweighted) vehicles. The latter option is the

focus of this paper.

Not including impacts from construction and maintenance of transportation

infrastructure, over 85% of life cycle energy consumption and greenhouse gas emissions of

(internal combustion) cars and light duty trucks arise from use-phase fuel consumption (Sullivan

et al. 1998; ANL-Argonne National Laboratory n.d.; Chester and Horvath 2009). There are a

number of factors responsible for vehicle fuel consumption including friction between the

system and its environment (e.g., rolling resistance, aerodynamic and hydrodynamic drag),

power source and transmission inefficiencies, type of power source, usage patterns, and vehicle

mass (M – short tons). Vehicle mass M is the sum of vehicle curb mass (Mc) and payload mass

2
(Mpyld), which is payload and/or passengers. Gross vehicle mass, Mgv, is the maximum value that

M can have for a given vehicle and is specified by the manufacturer. Because the lightweighted

(reduced mass) version of a vehicle uses less fuel than its heavier counterpart over the same

distance, manufacturers, especially those making cars and light duty trucks, often consider

mass reductions as a way to reduce vehicle FC.

As part of ongoing efforts to reduce FC, numerous studies have been conducted on

vehicle lightweighting (e.g., Koffler and Rohde-Brandenburger 2010; (Keoleian and Sullivan

2012); Lewis, Kelly, and Keoleian 2014; Kim and Wallington 2016; Mayyas et al. 2017). These life

cycle assessment studies address mostly cars and LDTs with various powertrains, including

internal combustion engine (ICE) vehicles (both spark and compression ignited), hybrid electric

vehicles (HEV), battery electric vehicles (BEV), and plug-in electric vehicles (PHEV).

Either mathematical models or empirical FC vs. vehicle mass correlations are required to

assess the impact of mass reduction on a vehicle’s FC. There are sophisticated computer

modeling tools used by vehicle designers for simulating the impact of mass reductions on

vehicle performance factors such as FC, gradeability, and 0-60 acceleration time. All vehicle

manufacturers have such modeling tools in-house (e.g., CVSP – Corporate Vehicle Simulation

Program) and the public has the Autonomie tool (formerly the Powertrain Systems Analysis

Toolkit-PSAT) (ANL-Argonne National Laboratory n.d.). However, there are many cost, energy,

emissions, policy, and life cycle analysts within the regulatory, academic, NGO, and

environmental communities who lack access or the desire to use such tools due to their

proprietary nature, cost, or sophistication. They too have an interest in the FC, cost, and

emissions tradeoffs associated with a vehicle mass reduction and would benefit from an

3
alternative approach, especially one like we present here that familiarizes them with the simple

physics underlying the mass dependence of vehicle FC. Examples of previous studies that have

used simplified methods to evaluate the energy and emissions benefits of vehicle

lightweighting are provided in the review by (Kim and Wallington 2013a).

The mass dependence of vehicle FC is an important quantity for estimating the merit of

a vehicle lightweighting initiative. For cars and light duty trucks (LDTs), this quantity is termed

the fuel reduction value – FRV (Lynne 1998; Koffler and Rohde-Brandenburger 2010). FRV is

fundamentally dependent on three factors: vehicle energy efficiency, rolling resistance, and

duty cycle (speed vs. time profile). The greater part of FRV is determined by duty cycle for a

given class of vehicle, with energy efficiency and rolling resistance of different vehicles within a

class contributing variation about that FRV ((Nam and Giannelli 2005) Helms and Krӓck 2016).

Fortunately, physics-based models have been developed to quantify all the components of

vehicle fuel consumption listed above, including that of vehicle mass (Kim and Wallington

2013b), from which vehicle FC, FRV, and fuel intensity (Fint = FC / Mpyld, in gal/100 ton-mi) are

readily computed. In fact, the computer modeling tools mentioned above are physics-based

models and they incorporate many vehicle and duty cycle variables and account for

dependences among them. An additional development (Kim and Wallington 2013b) is the use

of vehicle certification data (USEPA n.d.) for estimating FRVs, thus making it possible to

compute FRVs for a wide range of cars and LDTs listed in government databases over many

years.

The physics-based equations developed for light duty vehicles also apply to all other

wheeled vehicles, though core parameters such as rolling resistance, aerodynamic drag, and

4
engine, transmission, and miscellaneous losses for these heavier duty vehicles are unique to

each of them. However, a few reports (Delorme et al. 2009; Nylund and Erkkilä 2005) provide

enough information to allow determination of FC and FRV values for Class 6 and Class 8 trucks,

city buses, and LDTs using a simplified form of the basic equations. Basic models have also been

developed for estimating FRVs of trains (Davis 1926) and aircraft (Spakovszky n.d.). However,

computing FC and FRV for ships is more complicated due to a number of factors. Alternatives to

simple models for ships include both numerical (American Bureau of Shipping 2013) and

regression (empirical) (Holtrop and Mennen 1982) methods.

Our purpose here is to develop and assemble a set of simple physics-based fuel

consumption equations, henceforth denoted as the FCE Set, for estimating FC, FRV and Fint

values of the following transportation modes operating in the U.S.: cars, LDT, medium and

heavy duty trucks, trains, buses, and commercial high altitude jet aircraft. Electric passenger

trains were excluded due their energy consumption being only 0.2% of the U.S. transportation

total. These equations are used to estimate the impact of two weight reduction scenarios on FC

and Fint for each transportation mode. Each equation contains an FRV value that is

representative of its transportation mode. The impact of the vehicle duty cycle and other

factors on FRV are discussed. Sources of data are identified for determining FRV and other

terms in the equations. Where sufficient data permit, we compare FRVs computed from

experimental and simulation sources for cars and heavy duty trucks. Due to the absence of an

analytical expression for mass dependence of FC for tankers and container ships, values of FC,

Fint, and FRVs developed here were taken from the literature. Analytical expressions for FC, Fint,

and FRV are available for high altitude commercial jet aircraft and are used to calculate values

5
presented here. Finally, computed FC and Fint values are compared to values from the

literature. Our FCE Set provides analysts an alternative approach for estimating fuel

consumption impacts of mass reduction across modes in transportation vehicles while also

providing a physics-based understanding of this relationship without the need for sophisticated

or costly computer models.

2. Methods

The metric of key interest here is FRV (in gal/100 ton-mi), which is formally defined as

FRV = FC/M (1)

All FRVs considered here are for mass changes only. Powertrain adaptations combined with a

mass change are not considered, though powertrain adaptation combined with a mass

reduction can as much as double the FC savings from a mass reduction alone.

In three cases we use the FRVs developed here for each transportation mode to

estimate the impact of a mass reduction on its FC and Fint. The cases are: C1 - the base case, no

mass change, C2 - a vehicle mass reduction, and C3 - a vehicle mass reduction matched by a

payload increase, resulting in no change in overall mass M. Figure 1 depicts these cases and

their associated fuel consumption and fuel intensity expressions.

6
Figure 1 Illustration of vehicle and payload mass change cases (base case C1, vehicle mass reduction case C2,
vehicle mass reduction with payload mass increase case C3), with associated fuel consumption (FC) and fuel
intensity (Fint) expressions

For wheeled vehicles (cars, LDTs, medium and heavy duty trucks, buses and trains) basic

physics expressions can be written for computing fuel used to overcome resistance to motion

due to mass, aerodynamic drag, tire/road/rail losses, thermodynamic losses, accessory losses,

and internal powertrain frictional losses (Nam and Giannelli 2005; Thomas and Ross 1997; Kim

and Wallington 2013b). See Eqs. (A.1) to (A.6) and a brief discussion in Appendix A. Upon

inspection of Eqs. (A.1) – (A.6), it is clear that vehicle FC can be written as:

FC = FRV ∙ M +  (2)

where  denotes the sum of mass independent terms in Eq. 2. From Eq. S2 we have

FRV = FM / (M ∙ Dst) , (3)

where Dst is drive cycle distance and FM (the fuel associated with moving vehicle mass) is

defined as:

, (4)

7
where Hf is the heat content of the fuel, i is engine efficiency, t is drivetrain efficiency, A and

B are coast-down coefficients, a is acceleration and v is velocity. Details on these coefficients

and their application for determining FC components are given in Appendix A.

As drive cycles for wheeled vehicles tend to have short Dst, M remains sensibly constant

even though some fuel is being consumed and the vehicle is getting lighter. Eq. (2) is an

important relationship as it shows the vehicle FC for wheeled vehicles is linearly related to

vehicle mass, a relationship recognized by others (Kim and Wallington 2013b; Thomas and Ross

1997; Helms and Krӓck 2016; Nam and Giannelli 2005). However, to write equations for FC and

FRV in this way assumes that the mass changes considered are not sufficiently large enough to

affect engine efficiency or coefficient of rolling resistance. The computer simulation models

discussed above can address such effects, whereas Eqs. (2) and (3) are simplified physics-based

expressions that are expected to be valid for small to modest vehicle mass changes.

Reducing the mass of vehicles reduces both fuel consumption and transportation costs.

In addition to the voluminous amount of research focused on mass reduction in cars and LDTs,

there has also been research devoted to evaluating lightweighting of shipping containers

(Buchanan et al. 2018), container and tanker shipping (American Bureau of Shipping 2013),

trucks (Galos et al. 2015), and aircraft (Immarigeon et al. 1995)Calado, Leite, and Silva 2018).

FRVs are an extremely useful tool in estimating the benefits of weight reductions in all of these

situations.

2.1 Cars, Pickup Trucks (LDT), Medium and Heavy Duty Trucks, and Buses

Eqs. (2) and (3) apply to all wheeled vehicles, including cars, LDTs, medium and heavy

duty trucks, buses, and trains. In cases where an analyst does not have explicit values for the FM

8
term in Eq. (3), FRV can be determined from either experimental or computer simulation FC vs.

M data. In fact, for cars and LDTs the EPA vehicle certification database (USEPA n.d.) contains

extensive experimental vehicle data on many models over many years, for city and highway

drive cycles (speed vs. time profiles). For FRV estimates based on those EPA data, the most

relevant values are fuel economies (FE) for the city and highway drive cycles (FC = 1 /FE),

engineering test weight (ETW), and a set of three coefficients (A, B, and C) determined from

coast-down tests.

The three key expressions for estimating vehicle FC and FRV are Eqs. (2), (3) and (4).

Whether from certification data, experimental FC vs. M tests, or computer simulation output,

the coefficients in Eq. (2) (i.e., FRV and  are readily determined. In practice, for cars and LDTs,

FRV determinations are well developed using either experimental (Kim et al. 2015; Kim and

Wallington 2013b) or computational (Pagerit, Sharer, and Rousseau 2006) data. On the other

hand, for buses, trucks, and rail vehicles, the primary source of data is computational, though a

limited amount of experimental data are available.

2.2 Trains

Like with cars, trucks, and buses, the fuel consumed in moving a train is due to a

combination of friction, acceleration, and gradient effects, though the latter is generally not

considered. The FC for trains is computed from:

FC = ∫ (ΣRi + ) ∙ v ∙ dt / {i ∙ t ∙ Hf ∙ Dst} (5)

where Ri is the running resistance of each car in the train, is acceleration, v is velocity, t is

time, i is engine efficiency, t is drivetrain efficiency, Hf is the heat content of the fuel, and Dst

is distance. In cases with or without acceleration (), FRVs are readily computed using Eq. (5)

9
with appropriate sets of computed FC vs. mass data. A general expression for the friction

component R (in lbf) is the Davis equation (Davis 1926):

R = A + B ∙ v + C ∙ v2 (6)

where A, B, and C are coefficients determined from coast-down tests and v is velocity.

According to AREMA (AREMA 2015), A represents journal bearing, rolling, and track resistances,

B accounts for flange friction, flange impact, wheel/rail rolling friction, and rail wave action, and

C represents headwind pressure, rear drag, skin resistance on train sides, turbulence between

cars, and additional friction at the rails due to side winds. More on these coefficients and their

determination is found in Appendix A.

Values for the Davis equation coefficients have been developed for a number of train

types ranging from freight to high speed passenger trains (Rochard and Schmid 2000;

Lukaszewicz 2007; Boschetti and Mariscotti 2012; Davis 1926). All of these coefficient sets are

based on coast-down tests (Davis 1926; Lukaszewicz 2007; Rochard and Schmid 2000).

2.3 Aircraft

Aircraft of interest here are high altitude commercial passenger and freight jet aircraft.

An expression for the change in aircraft weight vs. time while cruising at altitude is:

-t∙v/Ω
W(t) = W(0) ∙ e (7)

where W(0) is the weight of the plane at takeoff, t is flight time, v is cruising speed, and Ω is a

constant dependent on the aircraft. Aircraft weight is dependent on time while in flight due the

consumption of fuel, which at takeoff can be a significant of aircraft weight. Eq. (7) is known as

the Breguet range equation; details on it can be found in (Spakovszky n.d.). A major fraction of

a commercial aircraft’s weight at take-off is fuel, most of which is consumed during the flight.

10
The total weight of a plane is the sum of passenger, payload, aircraft structure, and fuel

weights. Eq. (7) applies for both cargo and passenger aircraft. Typical airspeed for commercial

passenger and freight jets is about 560 mph (Tennekes 2009), about 0.84 Mach.

Average fuel consumption (lbf/mi) corresponding to Eq. (7) over the flight is:

-Dst/Ω
FC = W(0) ∙ { (1 - e ) + altitude / (Hf ∙ ) + 0.5 ∙ v2 / (Hf ∙ ) } / Dst. (8)

Eq. (8) accounts for fuel consumed while cruising but also for extra fuel needed to accelerate to

flight speed and climb to altitude (excluding the fuel needed to lift and accelerate the extra fuel

required for this). The majority of fuel on long and medium range flights (≈97%) is consumed

while cruising. Based on Eq. (8) and omitting the kinetic and potential energy terms to reach

cruising speed and altitude, FRV is:

-Dst/Ω
FRV = 3.0x104 e / Ω. (9)

Notice that FRV is not a constant and is dependent on distance. This is consistent with aircraft

FC being non-linearly related to aircraft weight.

An approach for computing a plane’s Ω value is to use one example of its takeoff weight

W(0) and corresponding FC vs. Dst (= v ∙ t) datum from, for example, simulation results from the

PianoX model (Lissys Limited n.d.). With that Ω, Eq. (8) can be used to estimate the minimum

fuel needed for the cruising stage of that plane for any Dst. In practice, extra fuel is carried in

case of airport backups or flight diversions.

Based on the relationships above, it appears that aircraft fuel consumption can be

totally related to aircraft weight. While this may be true to good approximation, there are other

losses, such as those associated with engine frictional losses and accessory loads (maintaining

cabin pressurization, lighting, powering equipment, heat and air conditioning), as well as fuel

11
use associated with taxiing to and from the runway. Because most of the fuel is used during

cruising, we use the above equations to estimate aircraft FC and Fint, despite omitting these

other loads.

2.4 Ships

To develop analytical expressions for ship fuel consumption, one must know the

resistance a ship is subject to upon movement in water. Major contributors to this resistance

are hull/water interface friction, wave making and wave breaking resistance, resistance due to

appendages, and other factors. In the past, this was done using models in tow tanks and scaling

to full size. Now this is often done using computational fluid dynamics software. More

specifically, these computational methods are applied to numerically solving the Reynolds

average Navier-Stokes equations in discretized form at the water/hull interface (Tupper 2013).

Another approach uses statistical regression methods (Holtrop and Mennen 1982). Others (Lu,

Turan, and Boulougouris 2013; Kwon 2008) have extended this approach to account for added

resistances such as sea direction, and wave period and height. Unfortunately, none of these

approaches provides a simple relationship between ship FC and weight (displacement). Our

approach here is empirical, relying on data provided in a study of a set of tankers and container

ships of standard design where the influence of weight on fuel consumption was estimated

using a computational approach (American Bureau of Shipping 2013).

Data sources that we used for estimating FRVs of various transportation modes are

shown in Table 1, and discussion of the relevant duty cycles follows below.

Table 1. Sources of data used to compute FRVs for various modes of transportation
Transportation Mode Source of Data
Cars and LDTs Certification data, manufacturer simulations, literature
Medium Duty Trucks Simulation data

12
Heavy Duty Trucks Simulation data, dynamometer tests
City Buses Simulation data
Rail Appropriate set of Davis coefficients
Aircraft From Breguet equation and PianoX data
Container ships and Tankers Ship simulation results from computational fluid dynamics study

3. Results and Discussion

The vehicles covered in this study are U.S.-operated cars (sedans), LDTs (Class 2a and

Class 2b), medium duty Class 6 trucks, heavy duty long-haul Class 8 trucks, city buses, freight

trains, passenger and freight jet aircraft, and tankers and container ships. A key characteristic of

these vehicles is their mass. In most cases, we refer to vehicle mass; for aircraft we use vehicle

weight (M ∙ g). We also use weight for trains, due to key Davis coefficients in the literature

being based on weight instead of mass. In the following sections, we develop our set of FC

equations (FCE Set) for the modes of transportation addressed here, including vehicle system

details.

3.1 Cars and LDTs

The two most widely characterized classes of vehicles in terms of FC (or FE) and FRV are

cars and LDTs. Though vehicle manufacturers don’t routinely publish FRVs for their vehicles, as

pointed out in (Kim and Wallington 2013b), they annually report fuel efficiency, emissions, and

vehicle size, weight, and horsepower details for each vehicle model to the EPA (USEPA n.d.).

Some of these data are key factors in FRV determination, namely coast-down coefficients and

city and highway FE values. We have computed FRVs from Eqs. (3) and (4) using certification

data for cars and LDTs covering a period of 17 model years; see Tables A.2 and A.3 for details.

All FRVs included in those tables represent metro-highway values and assume 55% city driving

(FTP – federal test procedure) and 45% highway driving (HWFET – highway fuel economy test),

13
and use the i and t values from Table S1. Values for the integral in Eq. (4) (as well as the other

integral equations in the SI) for those duty cycles are given in Table A.4; data for their speed vs.

time traces are found in (USEPA n.d.) and are depicted as Figs. A.1 and A.2.

The vehicles listed in Tables A.2 and A.3 were selected to represent the range of car and

LDT manufacturers and models and to span the range of masses for each vehicle class. All have

spark-ignition engines. Based on these values for 42 cars and 42 LDTs, average FRVs for these

two vehicle classes are 0.0409 ± 0.0040 gal/100mi/100lbs for cars and 0.0411 ± 0.0038 for LDTs.

Figure 2 plots FRV vs. mass for those 84 vehicles by vehicle class, and it is evident that FRV has

neither a dependence on vehicle mass nor class. Figure 2 is fully consistent with and

corroborates findings first presented by (Kim and Wallington 2013b) in their Figure 3d. Their

average value for 106 vehicles is 0.040 gal/100lbm/100mi. Our average for 84 vehicles is

0.0410±0.0038 gal/100lbm/100mi.

The values shown in Figure 2 represent vehicles manufactured between 1999 and 2016.

However, in 2008, the city duty cycle was updated by extending the speed vs. time profile to

account for both hot and cold start running conditions. What changed was the addition onto

the end of the UDDS (urban dynamometer driving schedule) drive cycle a repeat of its first 503

seconds resulting in a new duty cycle termed FTP (see Figure A.1). Nevertheless, the overall

character (number of stop and go events per unit distance travelled) of the FTP duty cycle is

sensibly the same. In fact, for a typical vehicle we calculate virtually the same FRV for both duty

cycles using Eqs. (3) and (A.2).

The reason for the apparent equivalence of car and LDT FRVs becomes evident upon

inspection of Eqs. (3) and (A.2), since FRVs for both vehicle classes are linearly dependent on

14
mass and both classes are subject to the same EPA drive cycle. What variation in the FRVs that

is seen in Figure 2 (and Tables A.2 and A.3) can be attributed to a combination of test

measurement variations of the A and B coefficients and vehicle-to-vehicle variation of rolling

resistance and engine and drivetrain efficiencies, the latter due to manufacturer choices on

vehicle performance details.

0.06

0.05
FRV - gal/100mi/100lbm

0.04
Cars
0.03 LDTs
CVSP
0.02
PSAT 1

0.01 PSAT 2

0
0 2,000 4,000 6,000 8,000 10,000
Vehicle Mass - lbm

Figure 2: Certification-derived FRVs vs. vehicle mass for a range of cars and LDTs; CVSP are
results from a study conducted by one of the authors while at Ford for 4 cars and 1 LDT; PSAT 1
are from (Pagerit, Sharer, and Rousseau 2006) for 2 cars and 1 LDTs; PSAT 2 are from (Delorme et
al. 2009) for a Class 2b truck.

From inspection of Eqs. (A.1) to (A.6), FC and FRV are expected to be dependent on

vehicle duty cycle. A number of researchers (Helms and Krӓck 2016; Pagerit, Sharer, and

Rousseau 2006; Delorme et al. 2009) have quantified this effect across a number of gasoline

and diesel vehicles for the FTP drive cycle in the U.S., the NEDC (new European driving cycle) in

the E.U., the JP10-15 in Japan, and others. FRVs derived from certification data also show this

trend. By applying Eqs. (3) and (4) for city and highway drive cycles using 2006 Lincoln Town Car

certification data we find its FRV to be 0.047 gal/(100mi/100lbs) for the city drive cycle and

15
0.034 gal/(100mi/100lbs) for the highway drive cycle. Clearly, FRVcty > FRVhwy. As is evident in

Figure A.1 and A.2, the reason for this is the many accel/decel events for city driving vs.

relatively few for highway driving.

We have some access to FRVs computed from vehicle simulation models (e.g., PSAT,

CVSP, Autonomie) for cars and LDTs operated in the U.S. A set of results is shown in Figure 2.

Upon comparing those FRVs to the values we derived from coast-down data in Figure 2, the

simulation FRVs fall into the lower half of the range of the certification results.

Notice in Figure 2 that FRVs for both the PSAT and CVSP simulation data decrease for

heavier vehicles. Other simulation results show a similar trend for conventional ICEVs, despite

their FC vs. M dependence being sensibly linear over a range of 40% mass reductions (Pagerit,

Sharer, and Rousseau 2006). This decreasing trend of simulated FRVs vs. vehicle mass for

conventional vehicles has been attributed to a decrease in engine efficiency upon a weight

reduction with no change in the drivetrain (Pagerit, Sharer, and Rousseau 2006). More

specifically, this decrease is attributed to a shift in the engine operating point on its torque vs.

rpm map. Other simulation results for both gasoline and diesel vehicles also show a

dependence of FRV on vehicle mass (Helms and Krӓck 2016), but in that case the trend is not

monotonic and those authors also note that average FRVs in the literature show no clear trend

with vehicle weight. Certainly, no dependence of FRV on M is evident in the certification-based

values in Figure 2. We expect FRVs derived from experimental results (e.g., certification data) to

be more accurate than simulation results, and we conclude that coast-down and simulation

FRVs are in adequate accord and that either approach provides reasonable estimates of FRVs

for cars and LDTs (excluding powertrain adaption after a weight reduction).

16
 Table 2 shows vehicle mass, payload mass, gross vehicle mass, duty cycle, FC, Fint, and

FRVs for the vehicle types considered here. The first two rows in the table contain data for a

specific sedan and LDT that are intended to be representative of those vehicle types.

Table 2. Vehicle specifics, fuel consumption, fuel intensity, and FRVs for a wide range of U.S. operated
vehicles; duty cycles are specified – combined cycle denotes either UDDS & HWFET or FTP-75 & HWFET.
a
Vehicle Mgv Mpyld FC Fint FRV Drive/Duty Cycles
short ton gal/100 mi gal /100 ton-mi
Cars and LDTs
Cadillac 2.25 0.15 2.1 3.69 24.6 0.788 combined cycle
Toyota Tundra 2.63 0.15 2.48 4.40 29.3 0.756 combined cycle
1
LDT-Class 2b 4.6 1.7 2.93 7.87 4.63 0.726 combined cycle
Heavier Duty Trucks and Buses
1 i
Class 6 Truck 13 8.1 4.93 17.8 2.20 0.812 HTUF P&D Class 6
1
Class 8 line haul 40 22.7 17.3 19.2 0.85 0.238 HHDDT65
1
Transit Bus 20.3 4.1 14.4 43.5 10.6 0.854 Manhattan
Rail
c
Freight train 7,175 4,100 3,074 746 0.18 0.045 40 CFR 1033.530
d
Freight train 7,175 4,100 3,074 1,150 0.28 0.105 “
Aircraft
e a g
Aircraft - 787-8 191 25.4 126.3 248 9.8 1.15 ADC
f a g
Aircraft – 747-400 346 69.8 197 500 7.2 1.26 ADC
Container Ships and Tankers
2
Oil Tanker 192,301 164,128 28,465 5,131 0.031 0.0290 unknown
2
Container ship 122,125 87,303 35,006 10,170 0.12 0.0714 unknown
a
This is the maximum operating mass as specified by the manufacturer often referred to gross vehicle mass, for
b
aircraft this is takeoff weight, for cars and LDTs this is taken as engineering test weight; This is the vehicle mass
without passengers or payload, which for vehicles on tires is curb weight (mass), for aircraft this is operational
c d e
empty weight; acceleration not included; acceleration included; Boeing 787-8, 242 passengers, no payload,
f
mass of passengers is 25.4 tons, 4000 mile flight, Ω = 21,800 miles; Boeing 747-400, 350 passengers, 30 tons of
g i
cargo, 4000 mile flight, Ω = 19,270 miles; our aircraft duty cycle, Figure A-5; HTUF Class 6 Parcel and Delivery
1 2
Truck; (Delorme et al. 2009); (American Bureau of Shipping 2013)

3.2 Class 2b LDT, Medium and Heavy Duty Trucks, and Buses

There is a dearth of simulation and experimental data for heavier duty vehicles,

including class 8 trucks. However, we found three sets of data for Class 8 trucks that permit

estimating additional FRVs, two from (Nylund and Erkkilä 2005), one denoted Freeway and the

other Highway, and the third from Argonne National Lab (ANL) (Delorme et al. 2009). All three

17
sets are plotted in Figure 3. Notice that the slopes of all three fit lines (equal to FRVs for each

data set) are virtually the same (about 0.24 gal/100 ton-mi). Because the dynamometer drive

cycles in (Nylund and Erkkilä 2005) are much like the drive cycle used by (Delorme et al. 2009)

(see Figure A.3), the similarity of these slopes is not surprising. All three drive cycles are mostly

conducted at constant speed with very few stop/go events. Another set of FRVs has also been

identified for heavy duty trucks (Helms and Krӓck 2016). Unfortunately, the duty cycles covered

there do not include HHDDT65 (heavy heavy-duty diesel truck), but the duty cycles included

yield FRVs that range from 0.18 gal/100 ton-mi for the WHVC (world harmonized vehicle cycle)

Extra Urban duty cycle to 0.31 gal/100 ton-mi for the HHDDT transient duty cycle. Our value,

0.24 gal/100 ton-mi, derived from (Delorme et al. 2009), falls at the middle of that range.

Figure 3 FC vs. Vehicle Mass for Heavy Duty Class 8 Trucks

Table 2 provides vehicle and fuel consumption details for heavy and medium duty

trucks, a light duty Class 2b truck, and a transit bus. The data for all four of those vehicle types

are derived from FC vs. vehicle mass data from the literature (Delorme et al. 2009). Based on

the agreement of FRVs, derived from both certification and simulation data across many cars

18
and LDTs, and based on the agreement between simulation and Finnish dynamometer FRVs for

heavy duty trucks shown in Figure 3, we conclude that the FRVs listed in Table 2 are

representative of wheeled vehicles in their respective classes operated in the U. S. Given that

vehicles within each class are designed to meet the same function and perform similarly during

their respective drive cycles, variations about their listed FCs and FRVs are likely due to small

variations in engine and transmission efficiencies and rolling resistance, much as seen in Figure

2 for cars and LDTs.

Table 2 also contains data for three additional road vehicles, namely a Class 2b LDT, a

bus, and a medium duty truck. For the Class 2b LDT, notice that its FRV is about the same as

that for the car and LDT listed above it. This a consequence of all three vehicles being tested to

the same duty cycle, despite the Class 2b LDT being much heavier. This is another

demonstration of the dependence of FRV on duty cycle. Notice also that FRV is around 0.85

gal/100 ton-mi for the city bus. When compared to an FRV of 0.66 gal/100 ton-mi from a

European report (Helms and Krӓck 2016), this value appears high. Actually, the two are in

reasonable agreement given the differences in their duty cycles. Though both duty cycles,

Manhattan and Braunschweig, are representative of bus stop and go profiles, the Manhattan

cycles has about two and a half times the number of stop and go events per mile as does the

Braunschweig duty cycle. Despite the former cycle topping out at around 25 mph and the latter

at 37 mph, the greater number of stop and go events for the Manhattan duty cycle is expected

to yield a somewhat higher FRV.

Upon inspection of Table 2, FRVs for these classes of road vehicles appear quite

variable. However, one must remember that these vehicles have different drive/duty cycles

19
that reflect how they are used in practice, which impacts their FRVs. The importance of drive

cycle characteristics was demonstrated above for sedans and LDTs. Hence, as seen in the table,

the Class 8 truck has lowest FRV, which is a result of a very low number of

acceleration/deceleration events, whereas the transit (city) bus has the highest FRV due to

many stop and start events. Their drive cycles are given in Figures A.3 and A.4.

Based on recent work for an all-electric bus (Bi et al. 2015), we compute an FRV of 0.474

gal/100ton-mi. This is roughly half of the FRV for a conventional diesel bus (Table 2), a trend

directionally consistent with FRVs for conventional and electrified (HEVs, PHEVs, and BEVs) cars

and LDTs (Kim and Wallington 2016).

Due to specific vehicle details and features within a vehicle class, one can expect some

variation in FRV. However, FC values might vary considerably between vehicles due to non-

mass dependent components of FC, e.g., aerodynamic drag or accessory loading. This is the

factor  seen in Eq. 2 and is easily computed with a representative FRV and a single FC vs. mass

datum.

3.3 Rail – Freight Train

To compute FC for trains requires both a duty cycle and expressions for train running

resistance, such as the Davis equation, Eq. 6. The train duty cycle that we employ (see Table

A.5) is specified by the EPA (USEPA 1997). However, we replaced the zero power ratings

specified there for dynamic braking and idle settings with 4.7% and 0.8% power, respectively.

We also assumed an accessory power value of 89.5 kW per locomotive, which is needed for air

compressors, fans, and auxiliary generators. Both of these changes were employed before

(Stodolsky et al. 1998). We also adopt their train, which is a freight train of 100 freight cars,

20
each weighing 17.5 short tons carrying two shipping containers each with a tare weight of 4.5

tons and holding 20.5 tons of cargo, and two 4500 kW locomotives each weighing 213 tons.

Train speeds given in Table A.5 were calculated from R ∙ v to match the required power at the

rails (% of max Engine Power ∙ i ∙ t) at each notch level. We assume i (brake thermal

efficiency) = 0.42 and t = 0.84. The procedure employed here to compute FRV is to calculate as

described above the speed consistent with the prescribed power (percent of total power) at

each notch level. We calculate an FC at the base weight of the train, then reduce the weight of

the train and, using the same velocities, compute a new FC. FRV is calculated from this pair of

FC vs. mass values.

Figure 4 Friction vs. speed traces computed with Davis coefficients for our 2-engine train pulling
100 cars of four different types.

Due to variation in published running resistance profiles for freight trains, we plot four

such responses all applied to the train described above. See Figure 4 for resistance force vs.

speed plots. They correspond to three North American Davis equation sets and one set based

21
on European coast-down data. All represent the freight train described above. One set

represents a train hauling container cars, another hauling flat cars with trailers on board, a

third hauling freight boxcars, and the last hauling both open and closed freight cars. See Tables

A.7 and A.8 for the coast-down coefficients (A, B, and C) used in Eqs. (5) and (6). Three of the

curves have similar slopes, albeit each with a different intercept. The fourth curve,

representing the modified Davis set of coefficients, has the lowest intercept but a much

steeper slope. Over the last 90 years, the coefficients for the Davis equation have changed due

to evolving equipment, changes from journal to roller bearings, changes in track design, and

changes in truck (wheel assemblies) designs, car configurations, higher speeds, and better

track (Hay 1982; AREMA 2015). Because the Canadian National Railroad set (flat cars with

trailers) is the most current and is recognized as providing reliable results in freight train

performance calculators (AREMA 2015), we adopt it as the set for our computations of freight

train FC, Fint, and FRV.

Two sets of train FC, Fint, and FRV are shown in Table 2, one with and the other without

acceleration effects included. Notice that values from the case where acceleration is included

are considerably higher than those from the case without acceleration, as expected. Because

the total distance travelled in the duty cycle is between 18 and 19 miles and most freight trains

travel much farther than that and at more or less constant speed, the best value lies between

the two and likely closer to the value without acceleration. Therefore, we chose the set without

acceleration effects as the most representative for freight trains operated in the U.S.

3.4 Commercial Jet Aircraft

22
 Pertaining to the FC of aircraft, our focus here is devoted to commercial jet airliners

used for transporting passengers, cargo, or both. See Figure A.5 for details on our aircraft duty

flight cycle. FC values for the aircraft shown in Table 2 were computed with Eq. 8 and include

fuel for climbing to altitude (PE - potential energy) and coming up to cruising speed (KE - kinetic

energy). Those PE and KE additions amount to 3% for long-range flight but are a higher

percentage for very short flights. FRVs were computed directly from Eq. (9). As indicated by

that equation, FRVs for these aircraft are dependent on distance traveled, unlike FRVs for road

vehicles.

3.5 Tankers and Container Ships

As mentioned above, we are unable at this time to write a simple expression for a ship’s

FC as a function of its mass (displacement). Nevertheless, we have a useful set of FRVs for

tankers and container ships based on a computer simulation study, seen in Table 3 (American

Bureau of Shipping 2013). These FRVs resulted from a set of FC vs. M values for replacing

small amounts of mild steel with high strength steel in the lightship. Notice in Table 3 that FRVs

appear to decrease with increasing design displacement for both classes of ships. Keep in mind

that these trends (depicted in Figure A.6) represent FRVs among ships of different

displacements and not FRV vs. displacement for a given ship. Also note that the data from

which these values were developed do not represent FC values due only to changes in mass.

In (American Bureau of Shipping 2013), the ship block coefficients were adjusted upon a

lightship weight reduction in order to maintain constant deadweight. Hence, these FRVs should

be considered only approximate. In general, reducing lightship weight is taken as an

opportunity to increase payload weight for volume-limited ships.

23
Table 3: Fuel consumption values, FRVs, and displacements for tankers and container ships
(American Bureau of Shipping 2013)

Ship Mdspl Mltshp Mpyld FC FRV Mltshp/Mdspl

short ton gal/100 gal/100
mi ton/mi
Tankers
Panamax 56,786 11,087 45,811 2,659 0.082 0.195
Aframax 133,466 21,299 112,431 4,004 0.039 0.160
Suezmax 192,301 28,478 164,203 5,131 0.029 0.148
VLCC 361,524 47,714 314,525 7,293 0.017 0.132
Container Ships
Feeder 17,928 5,539 12,416 1,926 0.112 0.31
Panamax 74,496 21,088 53,522 6,711 0.073 0.28
Neo-Panamax 76,296 21,035 55,377 7,333 0.090 0.28
Post Panamax 122,180 35,022 87,343 10,170 0.071 0.29
Ultra Large 183,371 51,910 131,739 13,057 0.061 0.28
Mdspl denotes displacement weight for a ship;
Mltshp denotes the weight of the ship not including freight or cargo (lightship).

Based on the data in Table 2, we now write specific FC expressions for each

transportation mode (with the exception of ships). Table 4 contains our FCE Set of FC equations

in terms of mass dependence for wheeled modes of transportation: cars and LDTs, Class 2b

LDT, Class 6 MDT, Class 8 HDT, a transit bus, and a train, as well as two aircraft. Each of those

expressions is unique to a particular example of each mode. However, for reasons discussed

above, FRVs for different examples within each mode are expected to be very similar, as we

showed for cars and LDTs across many manufacturers and a wide range of vehicle mass. Hence,

the FRVs listed in Tables 2 and 4 are expected to be representative for all vehicles within each

class. But because vehicles in any particular class or mode come in a range of shapes, different

size engines, with different engine maps, the intercepts of the FC equations are generally not

representative of all vehicles within that class or mode. Analysts interested in an FC equation

for a particular vehicle within a class need only apply the representative FRV and one

unadjusted FC vs. M datum to obtain the appropriate intercept () term in Eq. (2).

24
 Table 4. The FCE Set of fuel consumption equations for U.S. operated vehicles for seven
transportation modes
Vehicle Equation Mgv FC Units
tons gal/100mi
Cadillac FC = 1.92 + 0.788 ∙ M 2.25 “
Toyota Tundra FC = 2.41 + 0.756 ∙ M 2.63 “
LDT-Class 2b FC = 4.53 + 0.726 ∙ M 4.6 “
Class 6 Truck FC = 7.24 + 0.812 ∙ M 13.0 “
Class 8 Truck FC = 9.71 + 0.238 ∙ M 40 “
Transit Bus FC = 19.7 + 0.854 ∙ M 20.3 “
a
Freight Train FC = 424 + 0.045 ∙ M 7174 “
b -Dst/Ω c
747-400 FC = A ∙{ W(0) ∙ (1 - e )+ FPE + FKE } / Dst 346 “
c
787-8 “ 191 “
a b
Our freight train; Average fuel consumption over the flight; W(0) or takeoff weight; FPE = B ∙ W(0) ∙ Altitude;
2 -7 -9 2
FKE = C ∙ W(0) ∙ v ; A = 299 gal/ton fuel; B = 1.39x10 ton fuel/ft-ton aircraft; C = 4.64x10 tons fuel/mph /ton
aircraft, and overall engine efficiency eng = 0.5; Flight cycle in Figure A.5; Dst = 4000 miles; altitude = 35,000ft;
speed = 560 mph

3.6 Impacts of Mass Reductions

We now explore the impact of a vehicle weight reduction on FC and Fint for the vehicles

considered. Based on our FCE Set, we first comput the mass fraction (fcM = FRV* Mgv / FC) of FC

over their respective duty cycle. These numbers range between 0.4 to 0.6 for wheeled vehicles;

for the 747-400 this value approaches unity; fcM for ships can range from 0.8 to 1.8 for tankers

(decreasing with increasing displacement) and from 0.85 to 1.0 for container ships. The tanker

trend suggests that their FRVs are non-linear with displacement whereas FRVs appear roughly

constant with displacement of container ships.

The potential magnitude of an FC reduction is dependent on the vehicle’s FRV and the

unknown amount of a vehicle’s mass that can be reduced. An indication of the upper limit of

this potential is revealed in the ratio of vehicle mass (Mc) over Mgv. The higher this ratio is, the

higher is the potential for a mass reduction. Figure 5 presents a set of these ratios for the

modes of transportation studied here. Sedans and LDTs have the greatest potential for a weight

reduction and hence reduction in FC. This is because the gross vehicle mass Mgv of cars and

25
LDTs is mostly vehicle and is not passengers and payload. In the case of heavy trucks and trains,

much of their Mgv is payload. There is opportunity to lightweight our high altitude jet aircraft to

some degree, though probably carbon fiber composites would be the material of choice as

much of the plane is already aluminum.

Figure 5 The ratio of vehicle mass Mc to gross vehicle mass Mgv for several transportation modes

Figure 6 presents the %FC attendant a 10% Mc reduction vs. Mc/Mgv for different

vehicles. For wheeled vehicles, the trend is positive with FC reductions ranging from 2 to 4%.

The vehicles with higher Mc/Mgv values (sedans, LDTs, and buses) have the highest percentage

FC reductions. Notice that the % FC reduction for the 747-400 is well above those for wheeled

vehicles. The reason for this is that the resistance to aircraft motion in the air is more strongly

dependent on weight than is the case for wheeled vehicles.

Values for tankers and container ships are also presented in Figure 5. FRVs used to

compute the FC are taken from Table 3. Notice that the %FC values drop off quickly with

26
decreasing Mc/Mgv. This is primarily due to appreciable decreases in FRV as Mgv (ship

displacement) increases; see Figure A.8.

6
747-400
5

LDT-2a Sedan
% FC Reduction

4
Container
3 Bus
LDT-2b
2
M&HDT & Train
1
Tankers
0
0 0.2 0.4 0.6 0.8 1
Mc / Mgv

Figure 6 Percent FC reduction vs. Mc/Mgv for a 10% reduction in vehicle mass (weight)

Application of our FCE Set (Table 4) permits one to estimate the impacts of vehicle mass

reductions on FC and Fint. Fint results for wheeled vehicles and a commercial aircraft are shown

in Figure 7. The three Fint values represent: a base case (C1); a 10% Mc reduction (C2); and a

10% Mc reduction offset by an increase in payload (C3). The latter is conducted at constant Mgv,

i.e.,

Mpyld = - Mc (17)

Upon inspection of Figure 7, the overall trends between the modes are as expected:

trucks and trains have the lowest Fint values, whereas buses and aircraft have the highest

values. All C2 and C3 values are reduced when compared to their respective base cases, some

more so than others. As expected, the largest reductions in Fint are for the C3 cases. Another

feature evident in Figure 7 is that C3 scenarios for LDT-Class 2b and the jet airliner are

27
significantly lower than their base case values. The reason for this is that when Mc is much

larger than Mpyld a 10% decrease in Mc can make a significant increase in Mpyld.

12
2.8
10
Fint-- gal/100 ton-mi

5.1
8
22.8
6 2.7 C1
14.7
4 C2
2.2 5.8 C3
2
2.2 7.5
1.6 6.6
0

Figure 7: Impact of mass reductions on Fuel Intensity values for three scenarios each on six
modes of transportation: C1 --base case values; C2 for a 10% reduction of Mc; C3 for a 10%
reduction in Mc offset with an equal mass increase in Mpyld, i.e. -Mc = Mpyld; Reductions (in
percent) from the base case listed over blue and red columns.

It is not always clear if a C3 scenario is viable for a particular mode of transportation. For

example, an Mc reduction for a bus clearly moves passengers using less fuel but infrastructure

constraints such as the size and height of buses, road widths, and bridge heights might not

permit expanding the volume of a bus upward to admit more passengers, though articulated

buses would permit some expansion. Because highway lane widths and underpass heights limit

trailer volume, Class 8 long haul trucks might only be able to take advantage of case C2 changes

when hauling volume-limited cargo. If the truck is hauling a bulk load, it may take advantage of

case C3 scenario (up to the capacity of the trailer).

For ships, we find that a 10% weight reduction in lightship yields small improvements in

Fint. For scenario C2, Fint reductions range from 1–3% for tankers and from 2–3% for container

28
ships. With the exception of buses and aircraft, these percentages are in the same range as

those for trucks, trains, and LDTs. On the other hand, for C3 scenarios tankers have

considerably smaller reductions in Fint than what is observed of LDTs, medium and heavy duty

trucks, trains, and aircraft whereas containers ships have about the same percentage of

improvement in Fint as its wheeled counterparts.

Figures 8 and 9 present a comparison of model and published FC and Fint values for five

modes of transportation: trains, heavy duty trucks, medium duty trucks, buses, light duty trucks-2b

(LDT), and jet aircraft. Included in that list are results of Horvath et al. (Facanha and Horvath 2007,

2006; Nahlik et al. 2015). They conducted total life cycle assessments; we use only their vehicle

operational stage results. For each transportation mode, the figures show a range of FC and Fint.

Some of the variation within each mode is due to differences in vehicle payload (which affects FC)

while other variation is due to differences between annual mode averages vs. specific vehicle

results. Overall, results computed using our FCE Set for each of the transportation modes are in

reasonable agreement with their respective ranges. See Table A.8 for source information and

additional details. Note that there is considerable variation in FC and Fint for buses. This is primarily

due to variable occupancy rate as well as to comparing national average data with those based on a

fixed drive cycle.

29
Figure 8 A comparison of FC values from our model and published sources for truck, bus, train, and
aircraft

Figure 9 A comparison of Fint values from our model and published sources for truck, bus, train, and
aircraft.

Notice in Figure 9 for trains that there are two FC values from our FCE Set, one pair for

friction effects only and the other pair for friction plus kinetic energy effects. Since the drive

cycle for trains is only 19 miles long, FC for a friction plus kinetic energy scenario is more

representative of short range hauling while FC for a friction-only scenario is more

representative of long distance hauling at constant speed. Due to the scale in Figure 9, rail Fint

are hard to distinguish, but the values are in Table A.8. Because U.S. freight transportation via

30
rail tends to be long distance, the most representative values for Fint and FC are expected to be

closer to the friction-only values. Also notice the expected trend where the FC and Fint values

for rail are considerably lower than values for heavy duty truck (see Table A.8).

Another trend meriting comment is the good accord between Fint and FC values

computed from our FCE Set and values computed using an Ω from PianoX (Lissys Limited n.d.)

for the two aircraft listed in Table A.8. The two planes are the Boeing 787-8 and the Airbus

A380-800F. Both models are strongly based on the Breguet equation for aircraft cruising, which

is typically much longer in distance than climbing and descending. Though PianoX is more

sophisticated than our simplified model, our model nevertheless provides representative FC

and Fint values for these aircraft on climb-cruise-descend flight cycles.

A useful application of our simplified approach is estimating where in the transportation

sector vehicle mass reductions have the greatest impact, as illustrated in Figure 10. We assume

that our FRVs are representative and apply to all vehicles within each mode, a 20% reduction in

vehicle mass is realizable, the duty cycles assumed for each mode are typical of their use, and

all wheeled vehicles are powered by internal combustion engines, either spark or compression

ignited (hybrids and EVs are not included).

31
 Air
Rail
Buses
HDT
MDT
LDT1
Cars

0 2 4 6 8 10
Energy (TJ)
Friction_Accessories Drag_friction_Accessories
Due to current vehicle mass Energy saved due to weight reduction

Figure 10 Energy use by mode (in TJ) resulting from a 20% reduction in vehicle mass M c.

For the aircraft, note that the Friction_Accessories component represents engine

frictional losses and accessory loading due to heating, cabin pressurization, and electrical

generators for operating equipment and control surfaces. Aerodynamic drag losses for aircraft

are not included in the Fricition_Accessories bar as they are for wheeled vehicles. Instead, they

are incorporated in the weight dependent bar segment since aerodynamic drag is related to

aircraft weight (engine thrust must provide sufficient speed to overcome both aerodynamic

drag and provide enough lift (balancing aircraft weight) to keep the aircraft airborne). Figure 10

shows that, across transportation modes in the U.S., a 20% vehicle mass reduction results in the

greatest reduction in energy use for the LDT and car modes.

4. Conclusions

We have developed and assembled a set of fuel consumption equations (FCE-Set) derived

from general physics expressions for determining FC, Fint, and FRV for cars, light, medium, and

heavy duty trucks, trains, buses, and commercial jet aircraft operating in the U.S. Similar

32
expressions could not be developed for container ships and tankers. These equations are useful to

those who not only want to generate their own estimates of fuel metrics, but who also desire to

have a physics-based understanding of their derivation and representativeness. For wheeled

vehicles, these FC equations are linear in form but for aircraft they take an exponential form. Also

provided here are approaches and sources of data to determine the coefficients of these

expressions.

We have shown that FRVs for a wide range of cars and light duty trucks are remarkably

similar, despite significant differences in vehicle class, shape, and mass. This finding, which

corroborates previous results, demonstrates the dominance of duty cycle characteristic (city vs.

highway) on FRV. Observed Variations in FRV are attributed to vehicle-to-vehicle variations in

efficiency and rolling resistance and measurement error. Good correlation was found between

simulation and certification FRV estimates for cars and LDTs. Good correlation of FRVs for heavy

duty trucks was found for three separate studies, one a simulation study, one based on

dynamometer results, and a third European value. Based on the linear dependence of vehicle FC on

mass and the dominance of duty cycle on FRV, it follows that consistency between computed and

measured FRVs should be observed for all road vehicles within their respective classes when tested

according to mode-specific duty cycles. This is not surprising considering that vehicles within each

class (heavy duty trucks, medium duty trucks, etc.) are designed and built for the same function.

However, when comparing FRVs from vehicles with different powertrain technologies, such as

conventional vs. hybrid, such consistency between FRV is not expected. Hence, we conclude that

well-conducted simulation studies and/or experimental tests permit the determination of reliable

linear equations for FC and Fint for any wheeled vehicle.

33
 Using our FCE Set we estimated the impact of 10% vehicle weight reduction (not including

payload) on FC and Fint. For light, medium, and heavy duty trucks and trains, around a 2%

improvement in fuel consumption and intensity is realized. Values from 4% to 7% are estimated for

buses and planes, respectively. However, when a 10% vehicle weight reduction is offset by an

increase in payload weight at constant fuel consumption, the improvement in fuel intensity ranges

from 6% to 23%, the largest increase being for planes. Based on FRVs derived from simulations, a

10% weight reduction for ships can yield a few percent improved Fint for tankers, but a greater

percentage for volume-limited container ships.

Significantly, our FRV expressions permit reliable determinations of weight reduction on FC

and Fint. Though Fint and FRV have the same units (gallons/100 ton-mi), Fint is not a good estimate of

the impact of a vehicle weight reduction on FC. Fint is an allocated metric; FRV is a physical one. For

example, as seen in Table 2, Fint is about four times larger than the corresponding FRV for a heavy

duty truck, and hence using the former would result in a gross overestimate of the benefit of a

vehicle weight reduction.

Using the methods developed here to estimate the impacts of a vehicle weight reduction on

its FC, fuel and energy analysts in the freight transportation sector are now able to develop

reasonable estimates and evaluate the merits of weight reduction initiatives from both an

environmental and operational cost point of view. Shipping organizations would find our FRVs and

FCE Set of equations quite useful in estimating impacts of vehicle weight reduction on FC. Life cycle

analysts can use these estimates in combination with data on material production and vehicle

production, maintenance, and end-of-life stages, as well as life cycle models of transportation

infrastructure, to complete their life cycle assessments for each of these modes of transportation.

34
Finally, we used our FRVs to estimate the potential impact of lightweighting on transportation

energy use across the U.S. fleet of vehicles, across modes.

Acknowledgments

The authors gratefully acknowledge support for this work provided through Project Grant

Number F043665, sub-award agreement 0003 LCA from Direct Sponsor American Lightweight

Materials Manufacturing Innovation Institute and Prime Sponsor U.S.D.O.D. – Department of the

Navy.

35
References

American Bureau of Shipping. 2013. “Ship Energy Efficiency Measures: Status

and Guidance.”
http://ww2.eagle.org/content/dam/eagle/publications/2013/Energy
Efficiency.pdf.

ANL-Argonne National Laboratory. n.d. “Autonomie: Automotive System

Design.” Accessed March 14, 2017a.
https://www.anl.gov/technology/project/autonomie-automotive-system-
design.

———. n.d. “GREET2: The Greenhouse Gases, Regulated Emissions and Energy
Use in Transportation (GREET) Vehicle-Cycle Model.” Accessed February 20,
2017b. http://greet.es.anl.gov/greet_2_series.

AREMA, ed. 2015. “Manual for Railway Engineering.” In , 16-2-2 to 16-2–8.

American Railway Engineering and Maintenance of Way Association.

Bi, Z., L. Song, R. De Kleine, C.C. Mi, and G.A. Keoleian. 2015. “Plug-in vs. Wireless
Charging: Life Cycle Energy and Greenhouse Gas Emissions for an Electric
Bus System.” Applied Energy 146:11–19.

Boschetti, G., and A. Mariscotti. 2012. “The Parameters of Motion Mechanical

Equations As a Source of Uncertainty For Traction Systems Simulations.” In
XX IMEKO World Congress. Busan, Republic of Korea: Metrology for Green
Growth.

Buchanan, C.A., M. Charara, J.L. Sullivan, G.M. Lewis, and G.A. Keoleian. 2018.
“Lightweighting Shipping Containers: Life Cycle Impacts on Multimodal
Freight Transportation.” Transportation Research Part D: Transport and
Environment 62:418–32.

Bureau of Transportation Statistics. 2014a. “Table 4-12: Light Duty Vehicle, Long
Wheel Base Fuel Consumption and Travel.” In National Transportation
Statistics. United States Department of Transportation.
https://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/natio
nal_transportation_statistics/html/table_04_12.html.

———. 2014b. “Table 4-24: Energy Intensity of Transit Motor Buses.” In National
Transportation Statistics. United States Department of Transportation.
https://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/natio
nal_transportation_statistics/html/table_04_24.html.

Calado, Elcin Aleixo, Marco Leite, and Arlindo Silva. 2018. “Selecting Composite

36
 Materials Considering Cost and Environmental Impact in the Early Phases of
Aircraft Structure Design.” Journal of Cleaner Production 186:113–22.
https://doi.org/10.1016/j.jclepro.2018.02.048.

Chapin, C.E. 1981. “Road Load Measurement and Dynamometer Simulation

Using Coastdown Techniques - 810828.” In SAE Technical Paper. Society of
Automotive Engineers. https://doi.org/https://doi.org/10.4271/810828.

Chester, M V, and A Horvath. 2009. “Environmental Assessment of Passenger

Transportation Should Include Infraestructure and Supply Chains.”
Enviornmental Research Letters 4:1–8.
https://doi.org/http://dx.doi.org/10.1088/1748-9326/4/2/024008.

CSX. n.d. “Carbon Calculator.” Accessed October 24, 2016.

https://www.csx.com/index.cfm/responsibility/environment-and-
efficiency/carbon-calculator/.

Davis, William J. 1926. “Tractive Resistance of Electric Locomotives and Cars.”

General Electric Review 29 (10):685–707.

Delorme, Antoine, Dominik Karbowski, Ram Vijayagopal, and Phillip Sharer.

2009. “Evaluation of Fuel Consumption Potential of Medium and Heavy
Duty Vehicles through Modeling and Simulation.” Contract Number-DEPS-
BEES-001. Washington DC.

Facanha, Cristiano, and Arpad Horvath. 2006. “Environmental Assessment of

Freight Transportation in the U.S.” The International Journal of Life Cycle
Assessment 11 (4):229–39.

———. 2007. “Evaluation of Life-Cycle Air Emission Factors of Freight

Transportation.” Environmental Science & Technology 41 (20):7138–44.
https://doi.org/10.1021/es070989q.

Galos, J., M. Sutcliffe, D. Cebon, M. Piecyk, and P. Greening. 2015. “Reducing the
Energy Consumption of Heavy Goods Vehicles through the Application of
Lightweight Trailers: Fleet Case Studies.” Transportation Research Part D:
Transport and Environment 41:40–49.

Hay, William. W. 1982. Railroad Engineering. 2nd Ed. John Wiley.

Helms, Hinrich, and Jan Krӓck. 2016. “Enegy Savings by Light-Weighting - 2016
Update.” Heidelberg. http://www.world-
aluminium.org/media/filer_public/2017/05/02/ifeu_lightweighting_2016.p
df.

Holtrop, J., and G.G.J Mennen. 1982. “An Approximate Power Prediction
Method.” International Shipbuilding Progress 29 (335):166–70.

37
 https://doi.org/10.4271/971010.

Immarigeon, J-P, R.T. Holt, A.K. Koul, L. Zhao, W. Wallace, and J.C. Beddeos.
1995. “Lightweight Materials for Aircraft Applications.” Materials
Characterization 35:41–67.

Keoleian, Gregory A., and John L. Sullivan. 2012. “Materials Challenges and
Opportunities for Enhancing the Sustainability of Automobiles.” MRS
Bulletin 37 (4):365–73.
https://doi.org/https://doi.org/10.1557/mrs.2012.52.

Kim, Hyung Chul, and Timothy J. Wallington. 2013a. “Life-Cycle Energy and
Greenhouse Gas Emission Benefits of Lightweighting in Automobiles:
Review and Harmonization.” Environmental Science & Technology 47:6089–
97. https://doi.org/dx.doi.org/10.1021/es3042115.

———. 2013b. “Life Cycle Assessment of Vehicle Lightweighting: A Physics-Based

Model of Mass-Induced Fuel Consumption.” Environmental Science &
Technology 47 (24):14358–66. https://doi.org/10.1021/es402954w.

———. 2016. “Life Cycle Assessment of Vehicle Lightweighting: A Physics-Based

Model to Estimate Use-Phase Fuel Consumption of Electrified Vehicles.”
Environmental Science and Technology 50 (20):11226–33.
https://doi.org/10.1021/acs.est.6b02059.

Kim, Hyung Chul, Timothy J. Wallington, John L. Sullivan, and Gregory A.

Keoleian. 2015. “Life Cycle Assessment of Vehicle Lightweighting: Novel
Mathematical Methods to Estimate Use-Phase Fuel Consumption.”
Environmental Science & Technology 49 (16):10209–16.
https://doi.org/10.1021/acs.est.5b01655.

Koffler, Christoph, and Klaus Rohde-Brandenburger. 2010. “On the Calculation of

Fuel Savings through Lightweight Design in Automotive Life Cycle
Assessments.” International Journal of Life Cycle Assessment 15:128–35.
https://doi.org/10.1007/s11367-009-0127-z.

Kwon, Y.J. 2008. “Speed Loss Due to Added Resistance in Wind and Waves.” The
Naval Architect 3:14–16.

Lewis, A.M., J.C. Kelly, and G. Keoleian. 2014. “Vehicle Lightweighting vs.
Electrication: Life Cycle Energy and GHG Emissions Results for Diverse
Powertrain Vehicles.” Applied Energy 126:13–20.

Lissys Limited. n.d. “Piano-X: An Online Model to Estimate Commercial Jet

Aircraft Fuel Consumption an Emissions.” Accessed February 16, 2017.
http://www.piano.aero.

38
Lu, Ruihua, Osman Turan, and Evangelos Boulougouris. 2013. “Voyage
Optimisation: Prediction of Ship Specific Fuel Consumption for Energy
Efficient Shipping.” In 3rd International Conference on Technologies,
Operations, Logistics and Modelling for Low Carbon Shipping, 1–11. London,
United Kingdom: Naval Architecture, Ocean And Marine Engineering.

Lukaszewicz, Piotr. 2007. “Running Resistance - Results and Analysis of Full-Scale

Tests with Passenger and Freight Trains in Sweden.” Proceedings of the
Institution of Mechanical Engineers. Part F, Journal of Rail and Rapid Transit
221 (2):183–93. https://doi.org/10.1243/0954409jrrt89.

Lynne, Ridge. 1998. “EUCAR - Automotive LCA Guidelines - Phase 2 - 982185.” In

Total Life Cycle Conference and Exposition. Graz, Austria: SAE International.
https://doi.org/10.4271/982185.

Mayyas, A., M. Omar, M. Hayajneh, and A. R. Mayyas. 2017. “Vehicle Lightweight

Design vs. Electrification.” Journal of Cleaner Production 167:687–701.

Nahlik, Matthew J., Andrew T. Kaehr, Mikhail V. Chester, Arpad Horvath, and
Michael N. Taptich. 2015. “Goods Movement Life Cycle Assessment for
Greenhouse Gas Reduction Goals.” Journal of Industrial Ecology 20 (2):317–
28. https://doi.org/10.1111/jiec.12277.

Nam, Edward L., and Robert Giannelli. 2005. “Fuel Consumption Modeling of
Conventional and Advanced Technology Vehicles in the Physical Emission
Rate Estimator (PERE).” EPA420-P-0th ed. United States Environmental
Protection Agency: Office of Transportation and Air Quality.

Nylund, Nils-Olof, and Kimmo Erkkilä. 2005. “Heavy-Duty Truck Emissions and
Fuel Consumption Simulating Real-World Driving in Laboratory Conditions.”
In DEER Conference. Chicago, Illinois: VTT Processes.
https://www.energy.gov/sites/prod/files/2014/03/f9/2005_deer_erkkila.pd
f.

ORNL. 2009a. “Table 9.8 Summary Statistics for Class I Freight Railroads.”
Transportation Energy Data Book. 35th ed. Oak Ridge National Laboratory.

———. 2009b. “Tables 1.13 and 2.1.” In Transportation Energy Data Book, 35th
Ed. Oak Ridge National Laboratory.

Pagerit, S., P. Sharer, and A. Rousseau. 2006. “Fuel Economy Sensitivity to

Vehicle Mass for Advanced Vehicle Powertrains - 2006-01-0665.” In Society
of Automotive Engineers, 17.

Profillidis, V. A. 2000. Railway Engineering. 2nd Ed. Ashgate.

Rochard, B.P., and F. Schmid. 2000. “A Review of Methods to Measure and

39
 Calculate Train Resistances.” Proceedings of the Institution of Mechanical
Engineers, Part F: Journal of Rail and Rapid Transit 214 (4):185–99.
https://doi.org/https://doi.org/10.1243/0954409001531306.

Spakovszky, Z. S. n.d. “Unified Lecture #2: The Breguet Range Equation.” Edited
by D. Quattrochi. Unified: Thermodynamics and Propulsion. Accessed
January 1, 2016.
http://web.mit.edu/16.unified/www/FALL/Unified_Concepts/Breguet-
Range-U2-notes-Fall08(2).pdf.

Stodolsky, Frank, Linda Gaines, Roy Cuenca, and James J. Eberhardt. 1998.
“Lifecycle Analysis for Freight Transport - 982206.” In Total Life Cycle
Conference and Exposition. Graz, Austria: SAE International.
https://doi.org/10.4271/982206.

Sullivan, John L., Ronald Williams, Susan Yester, Elisa Coba-Flores, Scott Chubbs,
Steven Hentches, and Steven Pomper. 1998. “Life Cycle Inventory of a
Generic U.S. Family Sedan Overview of Results USCAR AMP Project -
982160.” In Total Life Cycle Conference and Exposition. Graz, Austria: SAE
International. https://doi.org/10.4271/982160.

Tennekes, Henk. 2009. The Simple Science of Flight: From Insects to Jumbo Jets.
Cambridge, Massachisetts: The MIT Press.

Thomas, Martin, and Marc Ross. 1997. “Development of Second-by-Second Fuel

Use and Emissions Models Based on an Early 1990s Composite Car -
971010.” In International Congress & Exposition. Detroit, Michigan: SAE
International.

Tupper, E.C. 2013. Introduction to Naval Architecture. 5th Ed. Elsevier.

USEPA. n.d. “Compliance and Fuel Economy Data for Vehicles and Engines.”
Accessed August 13, 2016a. https://www.epa.gov/compliance-and-fuel-
economy-data.

———. n.d. “EPA Vehicle Duty Cycles.” Accessed September 13, 2016b.
https://www.epa.gov/vehicle-and-fuel-emissions-testing/dynamometer-
drive-schedules.

———. 1997. “1033.530 Duty Cycles and Calculations for Railroad Locomotives.”
Locomotive Emission Standards.
https://yosemite.epa.gov/ee/epa/ria.nsf/vwAN/TS0000510E-
01.pdf/$file/TS0000510E-01.pdf.

USEPA, and USDOT. 2012. 2017 and Later Model Year Light-Duty Vehicle
Greenhouse Gas Emissions and Corporate Average Fuel Economy Standards,
40 CFR Parts 85, 86, and 600. https://www.gpo.gov/fdsys/pkg/FR-2012-10-

40
15/pdf/2012-21972.pdf.

41
 Appendix A

Equations of Motion for Wheeled Vehicles

For driving a wheeled vehicle over an arbitrary distance, the fuel used can be written as

the sum of five components:

F = ΣFi = FM + Faero + Facc + Ff + Fl (A.1)

where the values of the subscript “i” are M for mass dependent loadings (inertial and tire

losses), aero for aerodynamic resistance, acc for accessory loading, f for mechanical losses due

to engine friction and pumping, and l for other miscellaneous powertrain losses (outside of the

engine) such as in the transaxle, bearings, and differential. The units for F i are gallons or liters.

Each of the terms in Eq. (A.1) can be evaluated by integrating over a specified distance and set

of velocities a related power consumption expression for overcoming the each above cited

resistances to motion (Nam and Giannelli 2005; Thomas and Ross 1997; Kim and Wallington

2013b). The resulting expressions are Eq. (A.2) to (A.6).

(Eq A.2)

(Eq A.3)

(Eq A.4)

(Eq A.5)

(Eq A.6)

where a is vehicle acceleration, t is time, and ν is vehicle speed schedule corresponding to a

particular drive/duty cycle. The other symbols in these equations are defined with

42
representative values (Kim & Wallington, 2013; Kim et al., 2015) in Table A.1. With values for

specific vehicle constants, these equations can be used to estimate FC for any arbitrary duty

cycle (time vs. velocity profile), as per FC = FCi = Fi/Dst, where Dst (i.e., distance) = ʃ v dt and i

= M, aero, etc.

Table A.1: Definition of parameters in Eq. (A.2) - (A.6); representative values shown are
for light duty vehicles from (Kim & Wallington, 2013; Kim et al., 2015)
Symbol Definition Value Units
α Accessory power demand 0.00075 MJ/s
ε Rotational mass factor 0.1 -
ηi Thermal efficiency 0.41 -
ηt Transmission efficiency 0.88 -
ϕ Fraction of idling time 0.145 -
3
ρ Air density 1.2 kg/m
2
Af Frontal area 2.4 m
CD Aerodynamic drag coefficient 0.3 -
CR Rolling resistance coefficient 0.009 -
D Engine displacement 3 liters
fmep Friction mean effective pressure 0.150 MPa
2
g Acceleration of gravity 9.8 m/s
Hf Fuel low heating value (MJ/liter) 32.3 MJ/liter
M Vehicle mass 1500 kg
N Engine speed - rps
Varies over drive cycle
N’ Engine speed - RPM

Determination of Components of FC from Certification Data

During a coast-down test (Chapin 1981), a vehicle is taken up to a designated speed,

put in neutral and allowed to coast, during which the speed vs. time profile is recorded. From

this profile, an acceleration vs. speed trace is developed and fitted to a quadratic polynomial.

This trace corresponds to the force component of the instantaneous power response (P load) of

the vehicle during the coast-down test (Kim et al. 2015; Kim and Wallington 2013b):

(A.7)

where v is speed, a is acceleration, M is vehicle mass, and  is accessory load.

43
 The coefficients A, B, and C of the quadratic fit to the acceleration vs. speed trace

correspond to those in Eq. (A.7). Coefficient A is attributed to rolling resistance, B represents

higher order rolling resistance and mechanical rotational losses, and C corresponds to

aerodynamic losses (Chapin 1981; Nam and Giannelli 2005). During a coast-down test, the

parameters a (acceleration) and  (accessory losses) are zero. Because the engine during coast-

down is isolated from the powertrain on the road side of the clutch, engine friction Ff is not

included in the test data. Comparing Eq. (A.7) with the integrands in Eq. (A.2) and (A.3), we

write alternative expressions for those equations as:

, (A.8)

and

(A.9)

where i is engine efficiency, t is transmissions efficiency, t is time, and Hf is fuel lower heating

value (Kim et al. 2015).

A sample set of certification data is given in Tables A.2 and A.3. Tables A.2 and A.3

include vehicle specifics including coast-down coefficients (A and B), fuel economy, ETW,

and FRVs for the selection of cars and light trucks used for our analysis.

Table A.2: Vehicle specifics for cars between 1999 and 2015
a b
Year Model ETW FEcty FEhwy A B C FRV
2
lbs mpg mpg lbf lbf/mph lbf/mph gal/100mi/100lb
2006 Taurus 3625 22.1 34.6 27.26 0.548 0.0145 0.0433
1999 Grand Am 3375 22.1 35.6 33.64 0.00 0.0230 0.0373
2006 Ford Focus 3000 28.8 41.1 36.43 0.1963 0.0187 0.0452
2006 Lncln Town Car 4500 18.7 32.5 33.38 0.5859 0.0165 0.0418
2006 Toyota Camry 3625 25.8 41.5 34.43 0.0107 0.0207 0.0367
1999 GM Park Ave. 4000 20.5 36 41.42 0 0.0247 0.0379
2004 Chrysl Intrepid 3750 23.7 37.3 34.68 0.1806 0.0182 0.0393

44
2004 Ford Mustang 3625 21 34 14.61 0.9055 0.0133 0.0442
2005 Chrysl Neon 3000 28.1 41.3 31.03 0.5226 0.0151 0.0495
2003 Chrysl Intrepid 3875 20.9 34.9 43.59 0.1786 0.01978 0.0424
2003 Ford Crown Vic. 4500 17 27.8 45.83 0.72 0.0169 0.0483
2008 VW Jetta 3500 24.8 38.4 33 0.29 0.0171 0.0419
2003 Chrysl Sebring 3625 23.4 35.5 39.5 0.272 0.0182 0.0438
2005 Mazda 3 3000 28.5 43.2 28.0 0.276 0.0182 0.0424
2008 Malibu 3750 27.3 42.3 22.8 0.359 0.0191 0.0374
2003 Chrysler Neon 3000 28.1 41.3 31.03 0.523 0.0151 0.0480
1999 Kia Sephia 2875 25.4 40.1 38.27 0.235 0.0202 0.0466
2004 chevy Impala 3750 23.5 40.8 19.4 0.456 0.0166 0.0361
2005 VW Beetle 3250 24.5 39.9 30 0.150 0.0198 0.0376
2006 Mini-Cooper 2875 31.1 46.7 27.2 0.079 0.0185 0.0368
2008 GM Cobalt 3250 28.1 41.9 33.02 0.425 0.0164 0.0448
2009 Honda Civic 3250 25.6 40.5 30.67 0.249 0.0188 0.0400
2013 Taurus 4250 24.5 41.2 19 0.474 0.0201 0.0358
2015 Taurus 4250 24.3 39.5 23 0.491 0.0192 0.0375
2015 Chevy Malibu 4000 26 42 34.12 0.208 0.0191 0.0384
2015 Honda Fit 2875 42.6 54.7 35.1 -0.124 0.0219 0.0381
2013 Dodge Dart 3500 35.1 51.9 21.58 0.412 0.0149 0.0389
2013 Buick Regal 4000 23 40 33.58 0.493 0.0149 0.0429
2013 Chevy Caprice 4250 22.5 35.9 34.7 0.848 0.0150 0.0476
2010 Cadillac 4500 22.2 37 34.33 0.399 0.0193 0.0394
2012 BMW 3625 23.2 40.8 37.5 0.065 0.0184 0.0391
2012 Fiat 2750 35 46 35.53 -0.011 0.0218 0.0418
2013 Ford Fusion 3750 28.3 47.4 19.31 0.526 0.0137 0.0388
2010 Ford Fusion 3625 27.7 42.1 37.2 0.264 0.0168 0.0426
2010 Audi A3 3750 27.8 40 32 0.27 0.0200 0.0397
2010 Mini Cooper 2875 37 52 26.5 0.117 0.0176 0.0388
2012 Subaru Forester 3750 23.8 32.8 32.53 0.7948 0.0201 0.0492
2010 VW Jetta 3625 25.8 30.9 29 0.205 0.0168 0.0363
2012 Porsche 3500 23.7 38.2 31.47 0.110 0.0176 0.0363
2012 Ferrari 3875 14.4 24.4 44.06 0.199 0.0200 0.0415
2013 Toyota Corolla 3000 35.8 49.7 22.12 0.189 0.0181 0.0357
2015 Chyrsler Fiat 2625 38.8 53.3 23.49 0.244 0.0184 0.0403
2016 Audi ADX 3875 24.9 27.2 36.64 0.242 0.0173 0.0391
2013 Nissan Sentra 3125 34.9 50.8 19.84 0.291 0.0155 0.0361
a b
ETW is engineering test weight and for cars and light duty trucks it is curb weight + 300 lbs; metro-highway
values

Table A.3: Vehicle specifics for light duty trucks between 2003 and 2015
a b
Year Model ETW FEcty FEhwy A B C FRV
2
lbs mpg mpg lbf lbf/mph lbf/mph gal/100mi/100lb
2003 Chrysler Voyager 4500 14.1 22.1 32.43 0.740 0.0198 0.0437
2003 Ram 1500 5000 17.1 25.4 44.59 0.524 0.0346 0.0425
2003 F150 4750 17.9 26.1 17.219 1.390 0.0242 0.0464
2005 Toyota Sequoia 5500 17.1 23 46.738 0.506 0.0326 0.0410
2005 Fuji Forester 3500 24.7 35.9 33.29 0.681 0.0194 0.0494
2005 VW Touareg 5500 17.9 26.9 59 0.390 0.0296 0.0432

45
 2006 Chevy Suburban 6000 16.9 25.9 33.11 0.878 0.0284 0.0398
2006 GM Silverado 5250 12.8 18.6 26.89 1.659 0.0215 0.0504
2006 Ford F-150 5000 17.3 25.6 27.21 0.904 0.0288 0.0419
2009 Dodge Ram 5000 17.7 27.8 36.13 0.738 0.0300 0.0426
2009 Ford Expedition 6000 11.6 18.6 40.86 0.461 0.0360 0.0373
2009 Honda Odyssey 4750 19.7 32.4 31.77 0.821 0.0195 0.0434
2009 Chrysler Dakota 4500 18.1 28.1 47.55 0.204 0.0344 0.0413
2003 Chys Twn&Cntry 5000 17.9 29 40.9 0.7169 0.0213 0.0439
2006 Toyota Sienna 4750 20.3 30.6 47.4 0.1776 0.0294 0.0398
2009 Hyunda Santa Fe 4250 23 33.2 46.6 0.029 0.0290 0.0378
2008 BMW X5 5500 15.7 25.4 55.7 0.348 0.0270 0.0402
2006 Mercedes 5500 18.1 27 38.5 1.301 0.0120 0.0468
2005 Porsche Cayenne 5250 16.8 24.6 72.3 0.233 0.0299 0.0449
2004 Honda Element 3750 23.5 32.9 26.6 0.400 0.0272 0.0382
2003 Isuzu Axiom 4250 18.2 26.2 43.2 0.305 0.0321 0.0409
1999 Kia Sportage 3750 21.4 29.6 53.0 0.000 0.0316 0.0425
2010 Dodge Caravan 4750 19.7 31.9 35.94 0.651 0.02155 0.0425
2010 Dodge Ram 5500 16.6 26 46.99 0.602 0.03553 0.0422
2010 Lincoln Navigator 6500 15.3 24.1 51.43 0.896 0.03008 0.0431
2012 Chevy Silverado 5000 18.5 27.5 22.38 0.729 0.03141 0.0381
2012 Nissan Titan 5500 15.5 24.1 34.46 0.737 0.03385 0.0402
2012 Toyota Tundra 5250 19.8 27.7 28.947 0.616 0.03532 0.0378
2013 Ford F-150 4x4 6500 14.3 22.4 48.8 0.643 0.04637 0.0399
2013 Chevy Suburban 6500 12.7 21.7 48.31 1.071 0.02388 0.0441
2013 Toyota Sienna 4750 22.6 34.7 38.822 -0.047 0.02997 0.0337
2015 Ford Expl. FWD 5000 20.6 31.7 36.19 0.843 0.02253 0.0440
2015 Chevy Tahoe 6000 19.5 31.1 43.56 0.732 0.02549 0.0410
2015 Porsche Macan 5 4750 21.2 31.5 38.218 0.136 0.02824 0.0361
2010 Dodge Dakota 4750 18.1 28.1 47.6 0.2039 0.0344 0.0402
2013 Jeep Grnd Chrke 4750 18.5 27.5 38.1 0.8731 0.0257 0.0463
2015 Toyota Lexus 5000 36.9 38.8 38.3 0.2033 0.0264 0.0362
2010 Audi Touareg 5500 21.9 34.4 59.00 0.387 0.0296 0.0417
2012 Susuki Grand 3750 24.1 34.3 33.62 0.038 0.0256 0.0348
Vitara
2013 Mitsubishi 3750 28.9 39.7 45.45 -0.314 0.0276 0.0336
Outlander
2015 Buick Enclave 5500 20.6 33 46.76 0.540 0.0271 0.0399
2016 Dodge Durango 5500 17.5 30.8 53.68 -0.076 0.0309 0.0345
2016 Jeep Compass 3625 25.5 35.8 27.16 0.550 0.0232 0.0419
2010 Audi Touareg 5500 21.9 34.4 59.00 0.387 0.0296 0.0417
a b
ETW is engineering test weight and for cars and light duty trucks it is curb weight + 300 lbs; Metro-
highway values

Procedures for acquiring data in Tables A.2 and A.3

The sources of information needed for acquiring the data shown in Tables A.2 and A.3 are

located on the EPA website. The information comes from two Excel files. For the first file, go to

46
(USEPA n.d.) and click on “Annual certification test data for vehicles and engines”. Next click on

“Certified Vehicle Test Results Report Data (XLS)” for the model year of interest and download the

file, generally labelled in the following form: yearactrr.xls (e.g. 2016actrr.xls). This file contains not

only test track coast-down coefficients (choose target coefficients) but also vehicle details such as

engine displacement, number of cylinders, curb weight, GVWR, ETW, transmission details and

others. Fuel economy data are located in another file. For the second file, return to the “Annual

Certification Test Data for Vehicles and Engines” page, click on “Test Data for Annual Fuel Economy

Guide”, select the model year of interest, and download the data file found in the left most column.

It comes over as a zip file, e.g. 16data.zip. The zipped Excel file contains the fuel economy data for

the vehicles of interest. Of the fuel economies listed there, choose city, highway, and combined FE

unadjusted values that are consistent with test track results.

Values of the integrals used in calculating fuel economy and FRVs are given in Table A.4. For

computational convenience, we use SI units here; all other tables and figures displaying output

results employ imperial units.

Table A.4: Values of integrals used in Eqs. (A.2) - (A.6) corresponding to the FTP+HWFET
drive cycle.
Parameter Units City -FTP Highway - HWFET Metrohighway
2 2
ʃ v∙a∙dt m /s 3,026 1,151 2,182
ʃ v∙dt m 17,666 16,411 17,101
2 2
ʃ v ∙dt m /s 263,595 371,703 312,244
3 3 2
ʃ v ∙dt m /s 4,549,241 8,538,874 6,344,576
ʃ dt s 1,874 765 1,375

Figures A.1 to A.4 depict the duty cycles of various vehicles. Digitized representations of these

four duty cycles can be found in (USEPA n.d.).

47
Figure A.1: Duty cycle trace for light duty vehicle highway driving (HWFET – highway fuel economy
test)

Figure A.2: Duty cycle trace for light duty vehicle city driving (FTP – federal test procedure)

48
Figure A.3: Duty cycle (Manhattan) for buses: extracted from (Delorme et al. 2009)

Figure A.4: Duty cycle highway driving of heavy duty trucks; extracted from (Delorme et al.
2009), (HDDT65 - heavy heavy-duty diesel truck)

Coefficients for the Running Resistance of Trains

According to AREMA (AREMA 2015), the A, B, and C coefficients in the Davis equation

are defined as: A represents journal bearing, rolling, and track resistances, B accounts for flange

friction, flange impact, wheel/rail rolling friction, and rail wave action, and C represents

49
headwind pressure, rear drag, skin resistance on train sides, turbulence between cars, and

additional friction at the rails due to side winds. For any train, these coefficients are the sum of

the corresponding coefficients for each train unit, engines, and trailing cars. Overall, R for a

train is dependent on the mass of its cars and engines, number of cars and engines, train

length, number of axles, cross-sectional area of cars and engines, mass of payload, and other

factors. Based on extensive experimentation on train unit types and configurations, Davis (Davis

1926) developed explicit empirical formulas for these coefficients, which are:

A = F∙W + b∙n (A.10)

B = f∙W (A.11)

C = CA ∙ Af (A.12)

where F, b, f, and CA are constants, n is the number of axles per car, W is unit (locomotive or

car) weight in (short) tons, Af is cross-sectional area (ft2) of an engine or freight car, and v is

velocity in mph. Because these constants are historically based on train weight and not mass,

we write them in terms of weight W = M ∙ g. The constants are unique for each type of engine

and rail car, some examples of which are shown later. More detailed discussions on train

running resistance are found in (Profillidis 2000; Hay 1982).

Table A.5: Duty cycle in terms of specified notch duration and percent of rated power for train
fuel and emissions testing (USEPA 1997).
Throttle setting Duration (h) Estimated mph % of Rated Power
Dynamic Brake 0.125 12 0
Idle 0.380 0 0
Notch 1 0.065 8.0 4.5
Notch 2 0.065 14.3 11.5
Notch 3 0.052 28.1 23.5
Notch 4 0.044 35.6 35
Notch 5 0.038 43.3 48.5
Notch 6 0.039 51.4 64
Notch 7 0.030 57.7 85
Notch 8 at 63 mph 0.094 62.7 100

50
Table A.6: Equations for determining the Davis equation coefficients for freight trains from two European
sources; in SI units
2 2
Car Details A (N) B (N sec/m) C (N sec /m )
(Lukaszewicz 2007)
-4 -2
Covered cars 2000∙nloc + (65+6∙10 ∙ Q) ∙ nax -22 + 0.6 ∙ Lt 5.1 + 4.9 ∙10 ∙ Lt
-2
Open cars “ 0.3 ∙ Lt 5.1 + 9.2 ∙10 ∙ Lt
-2
Mixed cars “ -22 + 0.6 ∙ Lt 5.1 + 8.1 ∙10 ∙ Lt
nax is total number of car axles in train; Q is load per axle; Lt is length of train; nloc is number of
locomotives

Table A.7: Constants needed to determine the coefficients A, B, and C of the Davis Equation using
Eqs. (A.10) – (A.12).
Wheel Bearing Flange Aerodynamic
Friction Factors Friction Parameters
2
Car Type F b n f CA Af - ft
(Davis 1926)
Locomotive (leading) 1.3 29 4 0.03 0.0024 140
Locomotive (trailing) 1.3 29 4 0.03 0.00048 140
Freight cars 1.3 29 4 0.045 0.0005 90
Modified Davis – (AREMA 2015)
Locomotive (leading) 0.6 20 4 0.01 0.0024 140
Locomotive (trailing) 0.6 20 4 0.01 0.00048 140
Container cars 0.6 20 4 0.01 0.00104 90
Canadian National Railway – (AREMA 2015)
Locomotive (leading) 1.5 18 4 0.03 0.0024 160
Locomotive (trailing) 1.5 18 4 0.03 0.0006 160
Flat car with trailers 1.5 18 4 0.03 0.0005 125

Figure A.5: A comparison of assumed aircraft flight profiles (not to scale) for the models listed.

51
Figure A.6: FRVs for a set of 5 container ships and 4 tankers plotted at their design
displacement

Table A.8: Fuel consumption and fuel intensity values for five modes of transportation. Our model results listed in red.
Year FC Fint Mpyld Comment
Gal/mi Gal/100 ton-miles tons
Trains
1
1993 8.3 0.302 2,748 Annual average
1
1998 8.1 0.256 3,164 Annual average
2
2006 7.1 0.283 2,500 USA
3
2006 16.8 0.560 3,000 Calif.
4
1998 8.4 0.203 4,100 USA
5
2016 10.8 0.217 5,000 CSX calculator
6
2016 13.1 0.353 3,721 USA
6
2007 0.528 Calif
FCE set 11.5 0.280 4,100 with acceleration
FCE set 7.45 0.182 4,100 w/o acceleration
Class 8 Heavy Duty Trucks
Gal/100mi Gal/100 ton-miles %MaxPyld GVM/MaxPyld tons
7
2005 22.0 0.58 100 66/38
7
2005 15.5 0.55 100 46/28.1
2
2006 15.1 0.55 56 NA/49.2
8
2009 19.1 0.84 100 40/22.7
6
2007 17.6 1.41 56 NA/22.3
3
2016 2.49 Calif.
FCE set 16.3 1.43 50 40/22.7
Medium Duty Class 6 Trucks
Gal/100mi Gal/100 ton-miles %MaxPyld GVM/MaxPyld tons
6
2016 3.87
8
2009 17.8 2.20 100 13/8.1
FCE set 14.5 3.59 50 13/8.1
Buses
Gal/100mi Gal/100 ton-miles %MaxPyld GVM/MaxPyld tons

52
 8
2009 43.5 10.7 100 20.3/4.1
9
2014 28.7 Annual average
FCE set 40.9 20.4 50 20.3/4.1
LDT-Class 2b
Gal/100mi Gal/100 ton-miles %MaxPyld GVM/MaxPyld tons
8
2009 7.9 4.7 100 4.6/1.7
10
2014 5.8 3.7 1.6 Annual average
6
2007 2.8 1.8 100 NA/1.7
FCE set 7.3 8.6 50 4.6/1.7
Aircraft
Aircraft Gal/mi Gal/100 ton-miles tons Distance - mi
6
B747-400 9.4 70 Unknown
B747-400, FCE set 4.9 7.6 70 5,000
11
B787-8 2.4 9.6 25 2,000
B787-8, FCE set 2.2 8.8 25 2,000
11
A380-800F 8.4 5.1 165 6,500
A380-800F, FCE set 8.4 4.9 165 6,500
1 2 3 4 5 6
(ORNL 2009a); (Facanha and Horvath 2006); (Nahlik et al. 2015); (Stodolsky et al. 1998); (CSX n.d.); (Facanha and
7 8 9 10
Horvath 2007); (Nylund and Erkkilä 2005); (Delorme et al. 2009); (Bureau of Transportation Statistics 2014b);
11
(Bureau of Transportation Statistics 2014a); (Lissys Limited n.d.)

53