Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

802 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. IA-19, NO.

5, SEPTEMBER/OCTOBER 1983

The Interpretation of Train Rolling Resistance from


Fundamental Mechanics
STANLEY A. BERNSTEEN, RICHARD A. UHER, MEMBER, IEEE, AND JAMES P. ROMUALDI

Abstract-Past efforts in railway improvement have focused on etc.) can be classified according to its end use, i.e., inefficiencies
increased train stability and ride quality for safe and smooth opera-within the propulsion system, vehicle or train aerodynamic
tions at higher speed. The question of energy consumption has rarely
been considered. In view of projections of higher fuel costs and in-
resistance, dissipation of kinetic and potential (due to grades)
creased rail traffic volume, energy effects caused by improved train energy during braking, auxiliaries, and parasitic energy dissipa-
tion.
resistance should be considered. First-order expressions for the roll-
ing resistance of freight trains on level tangent track were derived Total energy consumption by any of these mechanisms is
from the basic principles of physics and engineering. The average extremely site specific and mode dependent. Reduction in
power dissipated by the suspension system was obtained from a linear
energy end use may lead to reduction of total energy con-
model of the track train system. Hysteresis in the soil is computed
using basic principles of soil mechanics. Losses due to wheel rollingsumption.
friction, imposed wheel sliding, and bearings were estimated from This paper is concerned with the problem of train rolling
simple engineering and considerations. Rolling resistance as would beresistance as an energy consumption end use. It is based in
measured in coast down or drawbar pull tests on level tangent track is
part on research carried out from 1976 to 1979 in much greater
found to depend strongly on the quality of the track as well as weight
and speed. On good track, most of the rolling resistance is due to
detail by the authors in [2].
bearing resistance, while on poor track, suspension losses predomi- Train resistance is generally defined as the net force oppos-
nate. Empirical expressions such as the Davis formula, which have ing the forward motion of a train exclusive of grades and
curves. Specific train resistance is the opposing force per unit
been used in the past by the railroad industry, are unable to describe
train rolling resistance adequately because of the way in which they gross weight, usually expressed in pounds per ton (lb/ton).
are parameterized. Some of the parameter values obtained from fits to
experimental data do not carry the physical meaning with which they
Specific train resistance is a dimensionless quantity-as such it
were traditionally associated. can be expressed as another dimensionless quantity-the
watthour per ton-mile (Wh/tonmi) which is the usual way to
express the specific energy consumption of a freight train. For
INTRODUCTION
example, the average energy, exclusive of propulsion system
AIL SHIPMENTS in the U.S. in 1978 reached 858 billion inefficiency, for all rail freight movement in the U.S. is about
tgross ton miles [1]. This freight was moved at an energy 40 Wh/ton-mi. This paper uses the latter unit (Wh/ton-mi) to
expenditure of 4.13 billion gallons of fuel at an estimated cost express specific energy consumption due to specific train re-
of 1.55 billion dollars (1978). In view of the rapid increase in sistance. The conversion factor is
fuel cost which occurred in 1979 and is expected to continue
into the 1980's and projected rail traffic volume increases lb/t 2(Wh/ton mi).=
(1)
which are expected to occur as a result of energy cost escala-
tion and tightening fuel supplies, programs for improving fuel An early comprehensive study of train resistance in the
economy which began in 1974-1975 are expected to continue U.S. was conducted by Davis in 1926. In his paper [3], an
and become even more popular. empirical expression for train resistance of the form
In the past, the primary concern of railway right-of-way im- B DX
provement programs have been to increase train stability and TR =A+ +Cv+ v (2)
ride quality in order to have safe and smooth operation at (W/n) W

higher speeds. The question of energy consumption effects is determined. The quantity TR is the specific train resistance,
which might arise because of changes in rail right-of-way was W is the weight, n is the number of axles, v is the speed of the
rarely considered. It was this initial consideration that led to train, and X is an effective frontal cross section. The constants
the beginning of the research to be described here. A, B, C, and D are empirically adjusted to fit the particular
Energy consumption in moving both freight and passengers, type of train considered.
both on rail and other types of guideways (trucks on roads, One experimental procedure to determine these empirical
constants is the coast-down test. The train is allowed to coast
This paper, approved by the Land Transportation Committee of on level tangent track while speed is recorded as a function
the IEEE Industry Applications Society for presentation at the 1980
ASME/IEEE Joint Railroad Conference, Montieal, Canada, April 9-10. of time. Deceleration is obtained by differentiation, and the
Manuscript released for publication June 9, 1982. retarding force is obtained by applying Newton's second law.
S. A. Bernsteen and R. A. Uher are with the Rail Systems Center, A second procedure, which generally has more uncertainty,
Mellon Institute, 4617 Winthrop Street, Pittsburgh, PA 15213.
J. P. Romualdi is with Carnegie-Mellon University, 4617 Winthrop involves measuring the drawbar pull required to maintain
Street, Pittsburgh, PA 15213. speed on level tangent track.

0093-9994/83/0900-0802$01.00 ( 1983 IEEE

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
BERNSTEEN et al.: TRAIN ROLLING RESISTANCE FROM FUNDAMENTAL MECHANICS 803

The accepted explanations of the components of the Davis material properties of wheel and rail, film deposits, state of
formula are that speed-independent terms account for bearing wheel and rail repair, and the dynamic load on the wheel.
resistance and sliding at the wheel-rail interface; the term Noise which is radiated as sonic and subsonic energy was
which is linear in speed is thought to be primarily due to estimated to contribute less than 0.13 Wh/ton-mi to total
flange resistance; the term which varies as the square of the rolling resistance at 60 mi/h where total rolling resistance is
speed is attributed to aerodynamic resistance, and the first typically 10-15 Wh/ton.mi. This estimate was based on the
three terms constitute train rolling resistance. worst case estimate and normally could be expected to be
Little theoretical justification exists for either the choice orders of magnitude smaller.
of (1) as the function describing train resistance or its inter-
pretation. One of the objectives of this paper is to show that Abnormal Rolling Resistance
indeed both the form of the function and its interpretation Hunting, flanging, unequal wheel radii on the same axle,
should be changed. However, the primary thrust of this 'and wheel lift are the most important forms of abnormal
research effort has been the viewpoint that energy of train rolling resistance. Parasitic energy dissipation in the draft
movement is dissipated partially by hysteresis in subgrade gear caused by grade changes and train handling is another
movement and partially by rocking and rolling of the vehicles form of abnormal rolling resistance.
which dissipate energy through dampers. The thesis is that Hunting, which is the severe lateral oscillation of the trucks
these energy dissipative activities can be modeled and energy arising from the interaction between the stiffness of opposing
consumption related to track condition, a hitherto unquanti- wheelsets, yaw, the gravitational restoring force, and the
fied effect. The established levels of energy dissipation due to lateral track profile, is a dangerous unstable condition which
these mechanisms are then compared to well-known dissi- truck designers try to avoid in normal operation.
pative mechanisms such as bearing friction and rolling fric- Hunting is usually accompanied by flanging, the banging
tion. Correction to existing rolling resistance formulas are also and rubbing of the flange of the wheels against the track when
made. the lateral displacement of the wheelset with respect to the
Another objective is to show that train rolling resistance track becomes too large. Wheel coning assures that flanging
depends strongly on track quality and, as a consequence, does not occur often.
energy savings as well as improved stability may be had with If wheels are out of round because of flat spots or other
good track maintenance programs. effects, if the lateral displacements are large, or if the radii of
the wheels on the same axle are different, the instantaneous
COMPONENTS OF TRAIN ROLLING RESISTANCE radii of the wheels at the wheel-rail contact area may differ.
Rolling resistance is the sum of the mechanical forces, This condition results in longitudinal wheel-rail slipping and
exclusive of windage and braking, acting to impede the for- wheelset yaw, which in turn can aggravate the hunting mode.
ward motion of a train traveling at constant speed on level Under severe dynamic conditions, a wheel may momentar-
track under operating conditions. These have been divided into ily lift off the rail or lose enough adhesion that lateral sliding
two parts which are classified "normal" and "abnormal." occurs. When the wheel restrikes the rail, either on the flange
or tread, nonlinear behavior will occur. All of these abnormal
Normal Rolling Resistance conditions generally lead to severe wheel wear and as such are
The energy loss components of normal rolling resistance are avoided by railroads whenever possible.
shown pictorially in Fig. 1. These include energy losses in the Another effect which is being considered along with ab-
suspension system, hysteresis in the soil beneath the tracks, normal train rolling resistance is the parasitic energy dissipa-
friction within the bearings, and sliding at the wheel-rail tion in the draft gear caused by train handling and rolling
interface. grade. Both of these are not considered in the rolling resistance
Suspension losses are a function of the vehicle design, track although they would be present in train operations.
quality, and the speed of the train. There is no apriori reason The abnormal effects are not estimated because they are
to assume that these effects should depend linearly on speed; assumed to be small on level tangent track where train resist-
however, it. is clear that if the train is stationary, no such ance is likely to be measured.
losses should occur.
Bearing resistance depends on the construction of the bear- DESCRIPTION OF MODEL
ing assembly, bearing dynamic load, lubricant type, and Suspension Energy Losses
temperature.
The amount of hysteresis in the ballast and soil depends on Both linear and nonlinear models of varying complexity
its properties and structure, tie dimensions, rail stiffness, and have been developed to approximate the car, track, and wheel-
the configuration and dynamic loading of the wheels of the rail reaction behavior in vehicle dynamics [4] -[10]. Linear
train. As in the case of suspension losses, speed dependence models are generally solved by straightforward numerical
is expected in this component. However, a static component techniques, whereas nonlinear models are handled by com-
(zero speed) must also occur. As the train passes over a section puter simulation. In most cases, the end result is a detailed
of track, the wheels alternately compress the soil and ballast description of the motion in terms of the coordinates, ride
underneath causing hysteresis cycles. quality indices, and ranges of conditions under which derail-
Sliding at the wheel-rail interface is determined by the ment is unlikely. Theoretical results are sufficiently close to

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
804 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. IA-19, NO. 5, SEPTEMBER/OCTOBER 1983

SOIL
(hysteresis)

contact area
-rail

IMEEEL-RAI L
INTERFACE
(rolling and
sliding friction)
Fig. .1. Sources cof rolling resistance.

experimental measurements to be considered reliable for into the wheels and forces at the wheel-rail interface. As the
engineering applications. wheels follow the rail, the forces at the interface represent the
A linear model was selected for the computation of suspen- acceleration forces due to the following motion plus the reac-
sion losses. This selection agrees with the philosophy behind tion of the vehicle suspension system.
the processes which cause normal rolling resistance, namely, The principal motions shown in Fig. 2 are
that in coast-down and drawbar pull tests to measure train * surface input responses of bounce and pitch;
resistance, coordinate displacements are relatively small and * cross level input responses of roll and twist (the influ-
that nonlinear abnormalities such as engaging suspension ence of gravity and wheel coning induce small lateral and
system stops do not frequently occur. Of course, this is veri- vertical motions which are found to be insignificant in
fied a posteriori by computing the probability for stop engage- terms of energy dissipation);
ment and flanging.
* alignment input responses of shift and yaw (wheel con-
Once selection of a linear model is made, Fourier analysis
methodology can be employed. Once the power dissipated in ing introduces small amounts of roll and twist which
are found to be insignificant in terms of energy dissipa-
various dampers is calculated by computing the time average
of the power developed by the sources, the average power can tion);
* gauge input responses of bounce and pitch due to wheel
be computed by summing the product of the Fourier ampli-
tudes of input coordinate speed and force over all input coning, which are also insignificant.
frequencies and coordinates. Thus the six principal motions to be modeled are bounce
To estimate inputs, track characteristics must be deter- and pitch caused by surface input, roll and twist caused by
mined. These are normally described as the symmetric and cross level input, and shift and yaw caused by alignment
antisymmetric combinations of rail functions which describe input. All other effects were shown to be small [2].
lateral and vertical displacements of the left and right rails Track profiles are usually measured by using a test vehicle
from their average displacement (zero on level tangent track) which senses vertical and lateral acceleration. Vehicle dynamic
as a function of the longitudinal position along the track. Each effects are minimized by suppressing the frequencies of in-
of the track characteristics, shown in Fig. 2, introduces motion terest [11]. The effect of vehicle weight on the track, ties,

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
BERNSTEEN et al.: TRAIN ROLLING RESISTANCE FROM FUNDAMENTAL MECHANICS

Track Function

r
y

F7.
I[/
:,,
z

s(y)=

(a)
bounce

__lA
= -s
=E _t
--l_

_I H
I_R
- \
ZR(Y)

B--- -zv bv
- - - - - --\ -

Fig. 2. Response of freight


s-
l- \
+

z
ZL(Y)

ballast, and subsoil cannot be eliminated, so that it would be


inappropriate to account for it in the model.1

tinuous welded rail, the deflections may be treated as an


ergodic stationary Gaussian random process. Regarding these
car

Fluctuations described by track characteristics arise because


of imperfections in installation, rail manufacture, etc. For con-

displacements as a stochastic process allows characterization of


the entire track rather than just the measured portion.
Track displacements are characterized in the literature by
their power spectral density (PSD) which, for "well-behaved"
track, is the density of the square of the absolute value of the
spatial Fourier transform of the track characteristic. 2 A range
L
c(y)

(b)
= ZR(Y)

K
- ZL(Y) a(y)

L
= XR(Y)

/ }

(d)

to four independent track functions. (a) Surface. (b) Cross level. (c) Alignment.
(d) Gauge.

Io-2

0-6
loor

01 .01
+
2

,-
XL(y)

(a)
x.v_-.S^_t
.1
-

g
small effect

g(y) = XR(y) - XL(y)

/.
/-
,5 X

L
{
/y w\ws
s-

/..

..
-
_
..I -{ \

f
,__
_.
_, _.__
t_ --

S(f)
\
-

\- X
=
o
X
\

g __ _ _t

=
R

-coD
s
S

Go

fs(y)eifydy
805

100
of PSD's are plotted in Fig. 3. The gauge function has little Q4
effect on vehicle dynamics and is generally not measured. 1lz8

Z5 1 C(f) fc(y) e fYdY


General agreement seems to exist among the forms of the =

PSD's but not their magnitudes, which are spread over more -2
than a factor of ten. This leads to the observation that con- 10

tinuously welded rail has basically one type of behavior multi-


plied by a constant factor which is a function of track quality.
Given a track function, the track class to which it belongs is
determined by finding its maximum deviation and comparing
it with the permissible deviations for the various track classes. .001 .01 .1 f
Given a Gaussian track with rms displacement a, there is theo- (b)
retically no limit to the magnitude of possible deviations, but " 1uu

displacements larger compared to 3c are extremely unlikely.


It is reasonable to assume that displacements exceeding three - 1
f0

times the rms displacement "never" occur. This cutoff is Alf) = J(f)eifYdy.
known as the 3a criterion. In particular, to construct the PSD _co

for a given class of track in accordance with the 3c criterion,

1 If unperturbed track functions were known, the problem would


be much more difficult to solve since it would involve an interactive .001 .01 .1
f
procedure involving track characteristics and vehicle motion. Because
perturbed track functions are used, this iteration is not required. (c)
2The curves in Fig. 3 are abstracted from [12], [21]-[24], [4], Fig. 3. Track power spectral densities. (a) Surface. (b) Cross level.
[7], [251. (c) Alignment.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
806 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. IA-19, NO. 5, SEPTEMBER/OCTOBER 1983

XOf (t)
HaL
(a)

qf(t)
Q (t)

(b)
Fig. 4. Decomposition of modified Batelle model into three independ-
ent models. (a) Surface. (b) Cross level. (c) Alignment.

adjust the constant factor so that a is one-third the maximum * capability of six motions of bounce, pitch, roll, twist,
allowable displacement for that class as determined by FRA yaw, and lateral displacements;
track standards. * numerical values must be available for mechanical springs,
The track data selected were supplied by Battelle [13]. It dashpots, masses, and dimensions;
was the most complete set of available PSD's and includes * reasonable simplicity and accuracy.
densities of jointed as well as continuously welded rail. These An exhaustive literature search indicated that the model
are shown in Fig. 3 as well. developed by Battelle comes closest to satisfying the criteria
Although the present method of reducing track displace- [12], [14]. The following simple modifications were made to
ments to PSD's causes the loss of phase information, the linear adapt the Battelle model to the purposes of this investigation.
model can still be completely solved. Because the track dis- 1) The rear half of the vehicle was modeled the same as the
placements are a stationary, ergodic process and the system is front. The addition of symmetry simplifies the model for this
linear, these PSD's can be integrated over the frequency to purpose.
yield the mean square of the response. Now, because the track 2) An effective lateral spring and damper between wheel
displacements form a Gaussian distribution whose mean is and rail were added according to the method of Cooperider
zero (because the mean of the excitation is zero) and whose [7]. This accounts for wheel-rail geometry and creep relation-
mean square has been determined, the probability distribu- ships.
tion of any response in the system is determined knowing The modified Battelle model, shown in Fig. 4, was restruc-
just the PSD's without the phase. tured in such a manner that either surface, cross level, or align-
The vehicle model must have the following characteristics: ment excitations could be considered by substitution of ap-
* linearity; propriate variables with a "generalized" format, as shown in
* independent (uncoupled) response to surface, cross level, Table I. The Appendix provides the solutions to the general-
and alignment excitations; ized model.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
BERNSTEEN et al.: TRAIN ROLLING RESISTANCE FROM FUNDAMENTAL MECHANICS 807
TABLE I
RELATIONS OF UNIFIED QUANTITIES TO THE SUBMODELSa

Unified Vertical Lateral CrcssIevel


MASSES _____5 MASSES RC' L t . S I_ERTIA
ml: axle IA
m2: track s%
r
in: carbody im
c

SPRINGS PR1CAS SRIGS TPIC.G SPRINIGS


k,c : wheel-rail k - k c k ' C
0 0 ~~~~~~~~zr'zr xr' xr ZT,r t;sr
k1c,c: primary suspension k
-..
c
zl
kxll cxlIi k c
x2' x2
CC2
k c: secondary suspension k~~~~~~ c -2
2' V ~ ~~ x2' ~ ~z2~C~z2' :;'
COUPLING MASS COUPLING MASS COUPLI:.; SPRINS COUPLING
A - ~~~~~~~~~~~~~~~~'~rtch
2 vyawic 24(
^ (c ,, tC2~~~~~~ ~~~~~~~
('Mc
aw2 4 (kc + ittc)
_2_) 12-2
COORDINATES COORDI!:TES CDC,PDI,3%TES COC NAATES
unified coordinates vertical displacement lateral' displace-ent roll angular displacemnent
TRACK EXCITATIONS EXCITATION EXCCITATION' EXCiTAION
Z(-a) : excitation PSD ' (u): surface profile A(w): aligrnent profile 9(w): cross'evel profile(anoular)
SOLUTIONS S 0oLUT:OIs SoLti,TIONs SOLUTIONS
P: power dissipation power dissipation power dissipation oower dissipation
Ff: force at front wheel vertical force at front lateral force at front torque in roll direction at front
wheelset: wheelset* wneelset**
Fr: force*at rear wheel vertical force at rear lateral force at rear torque in roll direction at rear
r wheelset' *heelseti wheelsetw*
Z: center of mass dis- b s
cm paeeni
pl-acement bonc bone|shif t roll
Zrel difference between pitch yaw twist
d isp lacements of
extremes of carbody
divide by the nuimer of-wheels per wheelset to get the force at one wheel.
divide by half the track gauge times the number of wheels per wheelset to get the vertical force at one wheel.
a See Fig. 4 for symbol definitions.

The modified Battelle model is a simplification of an actual because of the vehicle weight. Thus more motion is tolerated
freight car and is valid; according to Battelle, for input fre- in this direction and hence more dissipation occurs.
quencies up to 20 Hz. Above this value, nonlinearities become On continuously welded rail, energy consumption is ap-
important. This is sufficient, however, to account for the proximately a linear function of speed, whereas on jointed
major dynamic effects. Other limitations of the model, which rail power loss exhibits resonances. Above 25 mi/h trains
detract only slightly from its effectiveness, are running on jointed rail consume considerably more energy
than on welded rail of the same class. To find the correspond-
* substitution of a pair of axles for a single axle; ing values of energy dissipation for other classes of track, the
* invalidity below 10 mi/h because track spectral data at plots of Fig. 5 should be multiplied by the constants listed in
wavelengths less than a few feet cannot be used (lengths Table III as previously explained. The differences of order of
of the order of truck wheelbase); magnitude of the three contributions (surface, cross level,
* coupling between alignment deviations to lateral dis- alignment) to energy dissipation can be explained from the
placements is taken as constant, whereas it has super- model parameters (Table II) and deviations for class 4 track.
imposed upon it vertical dynamic forces;
* linear dashpots are idealizations of nonlinear damping Bearing Resistance
elements such as Coulomb snubbers.
In addition to suspension system effects, energy is lost to
The results of suspension energy dissipation are shown for bearing friction as well. Because of the trend toward the use of
class 4 track in Fig. 5 for welded and jointed rail. These were roller bearings by the railroad industry, only this type of
determined by using the PSD's for class 4 track as excitations bearing was considered in this analysis. Older friction bearings
in the frequency domain to the modified Battelle model and exhibit three to five times the friction of roller bearings.
computing the resulting average power dissipation. Parameter The primary causes of friction in roller bearings are elastic
values used in the model are tabulated in Table II. Surface hysteresis and slippage of the bearing elements. Conditions
inputs which result in bounce and pitch are found to be the which influence bearing resistance include the elastic proper-
dominant modes of energy dissipation. Cross level motions ties of its constituent parts, the loading, speed of rolling, and
are found to dissipate about one-tenth of this amount and the quantity and viscosity of the lubricant (which varies with
alignment even less. This would be expected from stability temperature). Experiments have been carried out to relate
considerations. Rail stability is highest in the vertical direction bearing friction to these factors empirically. For 100-ton

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
808

>: 0

44J

FE
12

is

1.2

1.e
/4

*
3/-~ ~ ~ ~Jointd
IEEE TRANSACTIONS ON INDUSTRY

2S
/01

so
APPLICATIONS, VOL. IA-19, NO. 5, SEPTEMBER/OCTOBER

velocity (mi/hr)
(a)
2g/

7s
b-~ ~ ~ ~ ~ ~ ~ ~Vlded

tSo
1983

bLJ
4-J2 .4
.2 / _ _, td

--Jointed
Welded

.02
* 25 Zs - 188

velocity (ini/hr)
(b)

.15

4 //
3 .5 z / -- Jointed

a
*s veIsocity (mi/hr" ISO
(c)
Fig. 5. Energy dissipation in suspension system (class 4). (a) Surface. (b) Cross level. (c) Alignment.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
BERNSTEEN et al.: TRAIN ROLLING RESISTANCE FROM FUNDAMENTAL MECHANICS 809

TABLE II
PARAMETER VALUES FOR THE MODIFIED BATTELLE MODEL 6

Parameter Vertical Crossieval | lceral 5

2 2 2
MASSES lb/finlsec) | lb/fi=/sec 3 lb/(im;sec-,
=1 - 7.59 7.51 7.59
°2 77.72 3.42 7.72 4
a 632.03 12. 632.03

SFlIOGS lblin 'lblin lb/iz


k 1 ,000,000 1,000,000 2 12 ci

k- g49,600 8.6,000 38,900


4H 3
k2 2,.100,000 766,700 rODO,000

DASI{PO0S lb/(ia/sec) 1b/(i*/sec) lb/(!=/sec)


co 2386 _385 613 ,676/v 2

c1 1824 31.62 1350


c2 100 II 2 2213

________
.^(
i_ _

lb/f!m/sec )
2
1f/gec 2 _- _

lb/(insece)
A 1 301.47 ~ 1067000 + 1110 2i 201.48 w

Other Valucs: len.gth of carbody: L - 1.68 inc hes


track gauge: C - 60 iccses 0 10 20 30 40 50 60 70 90 100

Scaled by a faczor of 1 , & - 60 incIes is tSh track puge. velocity, mi/hr


(C/2)
Fig. 6. Bearing resistance for trains of various weights (r = 18 in).
For conicity C.05 radians and y 1 (uo'orn vwsels)
v expressed in =les/hour. Coe-Fficien: of cz-ep is 2,700,000 lb.

TABLE III
CONVERSION FACTORS TO CONVERT CLASS 4 RESULTS TO
OTHER CLASSES

Class SUrface Zrcsslevel Ai;ment


"Powver ;LM5 Fr '.'S i LowS

6 X0..625 i 0.250 0.250 0.500 0.111 0.333


5 0.391 0.2.5 0.640 0.00 0.250 0.500
4 .0 1.000 .000 1300 1.000 1.00
3 1.266 1.125 w.60 1.400 1.361 1.167
2 3.357 1.833 2.941 1.715 6.618 2.572
1 4.000 2.000 6.618 2.572 18.38 4.287

RIS indicaces :te RLYS force or the i-S displacement.


Fig. 7. Model of right-of-way.
freight cars, roller bearing friction is given by
tainty in the soil parameters, any solution to the hysteresis
v1 /2 problem is just an estimate.
F= 202.4 3/2(W/n)2/3 Wh/ton mi -
(3) The model selected to represent the train-track system is
depicted in Fig. 7. The wheels are considered unsprung moving
where v is the forward speed of the car (mi/h), r is the radius loads, traveling along a pair of parallel rails supported by per-
of the wheels (in), W is the weight of the car (minus the weight fectly rigid ties. The ties rest up on the surface of a half-space
of the wheels and axles, tons), and n is the number of axles/ composed of a uniform isotropic medium whose response and
car. The tolerance for the expression shown in (3) is 20 per- hysteresis characteristics represent actual ballast material.
cent for speeds in excess of 0.3-0.5 mi/h. Hysteresis is computed by the three step procedure de-
The expression for energy dissipation of bearings does not picted in Fig. 8.
take into account the dynamic effects of loading caused by 1) From the configuration and weights of the wheels, stiff-
track inputs and suspension system response. However, it can ness of rails and tie dimensions, the pressure exerted by each
be shown that the time average fluctuations in the wheel load tie is found. The soil is modeled as a Winkler foundation [15].
do not affect (3) to first order. Fig. 6 shows bearing resistance 2) Given the pressure distribution upon the surface of the
as a function of speed and freight car weight. half-space, the stress is determined everywhere within the half-
space, which is assumed to be a linear elastic medium to which
Soil Hysteresis the Bousinesq solution applies [ 16] .

Computation of the energy dissipated in soil hysteresis due 3) As the train passes over a given point, the minimum and
to a moving train is complex. Because of the inherent uncer- maximum stresses at that point are computed. The hysteresis

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
810 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. IA-19, NO. 5, SEPTEMBER/OCTOBER 1983

w
W wt I

.z r 7L, I/ -7

_/

/1- -7----,

£)

c- L L L ,L L ~~
(c)
Fig. 8. Three-step solution to hysteresis problem. (a) Determine pressure exerted by each tie due to loading. (b) Deter-
mine stress beneath array of ties. (c) Determine hysteresis loss at each point and integrate over u.

loss per unit volume can then be found at each point using the The ballast itself does not dissipate an appreciable amount
stress/strain relations for ballast material. Integration over the of energy. However, soft ballast may allow the tracks to mis-
volume indicated gives the hysteresis loss per unit distance align and deform more severely than hard ballast. This will
along the track. induce more energy dissipation in the vehicle.
Using reasonable values of the parameters of the rails, ties,
track, soil, and load, the following conclusions are drawn. Rolling Friction
* The energy lost to soil hysteresis is estimated at 0.3 Rolling friction develops between the wheels and the rail
Wh/ton mi independent of train speed. Given the ap- because of wheel-rail contact. The frictional force opposing
the rolling of a cylinder on a plane is
proximations and the uncertainties in the parameters,
the estimate may be off by 100 percent. What is signifi-
cant is that energy loss due to soil hysteresis is small. Fx
=-1
r
(4)
* Most energy is dissipated within one foot of the surface
and near the edge of the ties. Little loss occurs more where r is the radius of the cylinder, N is the load, and X is an
than 8 ft to either side of the track centerline. experimentally determined coefficient. Thus the contribution
* Energy loss between 1-2-ft depths is 0.09 Wh/tonmi. to rolling resistance is
The loss density at those depths is nearly independent
whether the point is beneath a tie or the space between
the two ties. E=-Xr (5)
* In the neighborhood of the parameter values chosen,
increasing ballast resiliency or rail stiffness decreases and, for a railroad wheel (r = 18 in, X = 0.002-in steel on
energy loss. Heavier trains incur greater losses in the soil. steel), rolling friction dissipates 0.45 Wh/tonmi.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
BERNSTEEN et al.: TRAIN ROLLING RESISTANCE FROM FUNDAMENTAL MECHANICS 811

TABLE IV is explained by its quadratic dependence on amplitude of track


CONTRIBUTION TO TOTAL ROLLING RESISTANCE FROM ALL functions.
EFFECTS EXCEPT SUSPENSION LOSSES
Rolling resistance of full (120-ton cars) and empty (40-ton
Source irt cars) freight trains are compared for various classes of track in
Fig. 10. Curves for intermediate weights fall between these
Bearings 202.4 v ./2 extremes. Again, more energy is consumed on jointed than
r 3/J2 (f;/3
(W/n)
welded rail. Although it takes less energy to haul a lighter car,
Soil Hlysteresis 0.3
more specific energy is involved.
Ro1ling Friction 0.45 If a train consists of cars of weight Wi, for which the roll-
Imposed Wheel Sliding 0.25 ing resistance is Ei, the total resistance is
!ioise 0

total I.0 2 wh Eto=WiEi


WzEW (7)

v: speed of train (rpn)


W: weigant of car (tons) Comparison with Empirical Formulae
n: nurc.er of axles
r: wneel radius (in) Most of the empirical expressions for train resistance are of
the form of (2). The most widely accepted formulas are the
There is also an imposed wheel sliding because the wheels original Davis formulas (ODF) and modified Davis formula
have conical rather than cylindrical running surfaces. Since the (MDF). Table V lists Ahe coefficients for various formulas.
axis of the wheel is not parallel to the inclination axis of the This is not a demonstration of the experimental reproducibil-
cone, a rotational sliding occurs at the wheel-rail interface. ity of these coefficients, rather it was because these investi-
With the simplifying assumptions that the sliding occurs in the gators were primarily interested in determining D for different
contact area which is an ellipse whose length along the rail is types of rolling stock, the MDF or ODF values ofA, B, and C
about three times the width and that the force peak occurs at were assumed.
the geometrical center of the ellipse, the contribution to train The rolling resistance of a Battelle-Timken freight car on
resistance is class 63 welded rail is plotted in Fig. 11 for full, empty, and
partially loaded cars. The resistance predicted by model agrees
2 MaX
E=- (6) fairly well with the rolling portion of the MDF for heavy cars
9 r but not for lighter cars. This discrepancy can never be resolved
when empirical expressions in the form of (2) are used; the
where p is the coefficient of sliding friction, a is the cone slope of the curve is fixed in the empirical formula, but in
angle, r is the wheel radius, and X is one-half the major axis of reality it varies with the car body weight because of bearing
the contact ellipse. With ,u = 0.2, X = 0.5 in, ae = 1/20, andr = resistance and the suspension effects. On class 4 and 5 track,
18 in, the contribution to train resistance is approximately the MDF is inaccurate because its slope is independent of track
0.25 Wh/tonmi [17]. class, whereas the model predicts a dependence. This suggests
that the track where the test measurements which resulted in
DISCUSSION OF RESULTS the MDF were made might have been of class 6 quality. The
The contribution to total train rolling resistance from all Hoerner formula seems more appropriate on class 5 track, and
effects except suspension system energy loss is contained in the ODF seems to be applicable on class 4 track.
Table IV. All of the normal effects, with the exception of the
bearings and suspension, total 1.0 Wh/ton*mi. These effects Experimental Verification of the Model
have little weight dependence and are valid over the range of The coefficients of the Davis formulas and other such
weights considered (40-120 tons). empirical formulas were obtained by measuring train re-
Total Rolling Resistance of Freight Trains sistance over a range of speeds and train weights and fitting the
formulas to the data by a suitable selection of parameter
The rolling resistance of a Battelle-Timken freight train values. These experiments were performed in the presence of
(120-ton cars, 36-in wheels) with Timken roller bearings as air resistance, weather, and experimental uncertainties. How
computed using the modified Battelle model and accounting the weight of the car was varied is not clear; presumably all
for the other effects of soil hysteresis and wheel rolling fric- cars were tested under varying product load. However, dif-
tion is plotted as a function of speed for different track ferent type cars at lighter weight could have been used as well.
classes for welded and jointed rail in Fig. 9. In each case, the To eliminate unwanted effects, a test track of above quality
dotted line represents the rolling resistance on straight ideal was probably used.
track with no deviations in surface, cross level, alignment, A procedure which somewhat simulates the measurement
or gauge (no suspension losses), and the solid lines denote the of train resistance can be performed. Using the model pre-
boundaries between the indicated track class.
Suspension system losses are strongly dependent on track 3 Class 6 rail refers to the boundary which distinguishes class 5
class. The wide range of the magnitude of the suspension loss from class 6.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
00

12

11 L

10 10 F tiI
H
9
9 z
8
;,q
8 Class 4
E
1 .,, z
E

7 Ss 7
g: 3 z

6
4J
e4J
tn

.,4
_r 4>c
u C- s
1-

cx g:
z
-4
Class 6 CQ
Ideal Rail
3
3 cjT
0
0
Lf
to
H
cn

L. _1, I
20I
0 10 20 30 40 S0 60 70 P.O 90 100
). lO 20 30 40 590 60 70 80 90 100
Speed, mi./hr Speed, mi /hr
(a) (b)
00
Fig. 9. Rolling resistance of full freight train. (a) Welded rail. (b) Jointed rail. z
oA

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
z

is :~~~~~~~~~~~~z

7 *r
Empty Cars
(40 tons)

E s~~~~~~~~~~~~~~~~~~~~~~~* 10
EMDty Cars C
... 0
z~~~~~~~~~~~~~~(0 os
c~~~~~~~~~~~~~~~~~
r- 4
ar
C~~~~~~~~~~~~~~~~Fl
bo~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.1-4 Cl Cars
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1-4
~~~~~~~~~~~~~~~~~~~5*.'(120 tons)

2 V
...... Jointed Rail ......Jointed Rail-

WVelded Rail IlWelded Rail-

0 10 20 30 40 5o 60 70 80 90 1000 10 2 30 4 50 6 70 8
Speed, mi/hr Spe,iih
(a)(b
Fig. 1 0. Rolling resistance of freight trains. (a) Class 6 track. (b) Class 5 track. (c) Class 4 track. (d) Class 3 track.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
814 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. IA-19, NO. 5, SEPTEMBER/OCTOBER 1983

20 _ F,mpty rars
(40 tons)
18

16 16 FanDty Cars
(40 tons)
14 14

12 .,H
.: FE 12
Full Cars E2
(120 tons) $-1
u 10 Full Cars
co (120 tons)
~0
.H 8
n Cs
0
ce
6

_2

10 20 30 60 10 20 3
Speed. mi/hr Speed mi/hr
(c) (d)
Fig. 10. (continued)

TABLE V
COEFFICIENTS OF SOME EMPIRICAL FORMULAS: A + BvU [c/(W/n)] + D(v2/W)

E-mpirical Formula A B C D X 1000


20]
Ori.inal Davis Formula ] (plain 2.6 .090 58 130
bearings)
[21]
K;ranyi Form:ula 2.6 .090 58

i.ioerner Forrrula '2 2.8 .040 56 66.4


[20]
C'R >odified Davis Formula 1.2 .020 40 140
(roller bearings)
EL :dified Davis Formula
[20] 1.2 .020 40 400
(roller bearings)
(unconventional frcigh.)
t)
Dyer Formula 1.2 .020 40 140
[22]
British Formula 5.0 .097 0 1.034 W

**Frontal area assumed to be 130 ft .


Table for various types of rolling stock in Reference 22; value given for
conventional freight.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
BERNSTEEN et al.: TRAIN ROLLING RESISTANCE FROM FUNDAMENTAL MECHANICS 815

.E
.-
3R

ur

cY

=1
0)

Speed, mi/hr
Fig. 11. Comparison of model with modified Davis formula on Class 6 welded rail.

viously described, data points are computed at 1 mi/h intervals,


and in each case the empirical formula (2) was fit to the data
by the method of least squares. Four different weight trains TABLE VI
were used in the simulated tests. Table VI lists the results of FITS TO SIMULATED EXPERIMENTAL DATA UNDER VARIOUS
TRACK CONDITIONS
the experiments for certain track conditions and speed ranges.
The following conclusions are valid. Veloc.tL
Track
1) Rolling resistance on class 6 track agrees with the roll- (to) C..dit A IX c D x 1000

ing portion of the MDF; on class 5 track agrees with the roll- 10-80 lIdal .rack .93 + g4i .020 .018 25 1' .S .5
ing portion of the Hoerner formula; and on class 4 track 10-0 Clas 6 Weldd .94+1 2.22 .023 .024 26 14 1.1+ 1.1
with the rolling portion of the ODF. 10 ' Class 5 Welded .93 + 2.60 .060 + .051 is'30 3.4 + 2.3
2) Values for the weight-independent speed term (B) 1060 IC1ss 4 elded .80 + 2.55 .095 4 .065 27 29 8.0 + 3.9
agree very well in all of these cases, possibly indicating the 10-40 Class 3 '.;ld*d .67 + 1.80 .135 + .067 20 + 19 l.B 5.6
class of track on which tests were made. 10-80 jClas 6 Jointd .97+ 165 .023 .033 27 + 19 1.3- 1.5
3) The computed value of the speed and weight inde- 10-80 Class 5 jolated 102 +478 .046_ .094 35 + 56 4.5 +4.3
pendent term (A) fall consistently lower than the empirical 10-6 Clas1. Jointd '.2 6.48 .116 + .168 14 74 12.8 + 9.8
values by about 0.3 Wh/ton-mi. This may be caused by poor 1C-40 Clas 3 Jointed 4+ 593 .270± 220 8 63 12.U 18.4
estimates in the combined estimates of soil hysteresis, rolling
friction, and imposed wheel sliding. It could also be due to :...as n tr.a refers to track geonetry at t±a class a - clas (n ) oundary.
higher values of speed dependent terms in the suspension
losses for which no empirical fit terms were included.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
816 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. IA-19,NO. 5, SEPTEMBER/OCTOBER 1983

4) The values of the weight-dependent speed-independent would lead to expressions for rolling resistance applicable
term (C) fall below the empirical values. This is suspected to under a wider range of conditions.
be caused by air resistance which, as a combination of shape Curve resistance is empirically set at 1.6 Wh/ton*mi per
and skin effects, varies as V1 85, which when fitted to a term degree of curvature independent of speed. Clearly, dynamic
such as effects must be present with some speed dependence.
Air resistance has been neglected. At higher speeds, aero-
v V2 dynamic forces are larger than rolling resistance. The aero-
C ±D- (8) dynamic properties
(w/n) W of a train, as a function of types of rolling
stock and their order in the consist, are largely unknown.
would introduce positive values of C and D. Experiments have yielded estimates of air resistance for
5) The v2 term (D), which is usually interpreted as aero- some types of cars, but theoretical justification of the result-
dynamic in nature, is due in part (0-10 percent) to rolling ing empirical expressions is lacking. Much work remains, and
resistance especially for poorer track. There is dependence on a new interpretation of the experiments should be possible
higher powers of v as well, especially poor jointed rail. because of the work reported here.
6) The magnitude of uncertainties of the fit coefficients Theory is only as good as its agreement with experiment.
indicates that the fit is not good and that (1) does not describe Data must be obtained with cars of varying suspensions,
the computed curves well. This fit becomes more unreliable weights on different track quality under controlled conditions.
the lower the class of track. None of the empirical formulas An empirical formula which may predict train resistance
are valid for all car body weights or all classes of track. better is

SUMMARY V1/2 v2
+ C (r)3/2(W/ )2/3 D
TR =A+ BV± (9)
Conclusions nI W
This paper discussed the modeling of a freight train and its Finally, a more detailed examination of car behavior as a func-
rolling resistance. The consequence of computations made tion of suspension elements may lead to design of a suspension
with the model are as follows. which has lower energy dissipation, higher stability, and better
The modified Davis formula, which has gained acceptance ride quality.
in the railroad industry as a good empirical expression for train
resistance, has the best agreement for heavily loaded trains on APPENDIX
good quality track. It becomes poorer as the cars are less The total average power consumption is given by the ex-
loaded and the track quality decreases. pression
Some of the terms in the empirical formula do not have the
physical significance usually associated with them. Their co- P.=
efficients merely produce the best fit of the functional form
to the experimental data.
Rolling resistance depends strongly on class of track. Up- where the power at frequency X due to an excitation function
grading of railroad track, however, may not lead to better of the form
energy consumption, since higher track standards generally
lead to higher speed freight, which will, because of the addi- Z(o),)eiwt
tion of aerodynamic effects, negate some of the effects of the
rolling resistance improvements. However, more care in the is just
surface alignment can lead to better energy efficiency since
most of the suspension train resistance is due to this mode. P(w)= Co Z(w) 12 Im [H + A(1 - cos co)D D']
More energy is consumed via rolling resistance on jointed
rail than on welded rail of equivalent class. This has been sus- where
pected for some time, but the magnitude of the difference was ,(-m2co2 +K1 +K2)(-mco2 ±2K2)-2K22
not known. H=Ko -KD2 d
To first order soil hystersis in ballast beneath the track is
a speed-independent effect. However, it is important to keep
the ballast well maintained, for good ballast foundation helps D= KoKIK2
d
prevent deterioration of track geometry.
Dr _KOKIK2
Future Research dt
Since theoretical knowledge in the area of train resistance and where
is rather limited, further investigation could be of practical d = (-mw2 + 2K2)[(-m2CJ2 + K1 + K2)(-mIco2
value to the railroad industry. Refinement of the model and
inclusion of some of the more important nonlinear effects +Ko +K1)-K12J -2K22(-mIc o2 +Ko +K1)

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.
BERNSTEEN et al.: TRAIN ROLLING RESISTANCE FROM FUNDAMENTAL MECHANICS 817

d'= (-mico2 + A + 2K2) [(-m22'2 + K, + K2)(-m1c2 [18] D, Tope, "Rolling resistance of freight cars," Timken Bearing
Co., Columbus, OH.
+Ko +K)-K12] -2K22(-_Mn 2 +Ko +K1). [19] "Review of train resistance equation," Dep. Mech. Eng., Union
College, Oct. 1976.
The quantities m, mI, M2, Ko, KI, K2, A refer to Table [20] J. Muhlenberg, "Resistance of a freight train forward motion,"
MITRE paper MTR-7664.
I and Fig. 4. The following are defined further: [21] A. Gilchrist, "Power spectral measurements by TMMI: Proving
trials and three site measurements," British Rail Res. Dep., Sept.
Kj = kj + icj, complex spring constants, 1967.
v train's forward speed, [22] K. Schrotberger, "ORE Organization and its studies in track, sus-
w input driving frequence (c = 27ivf) where f is the track pensionsion and track-train interface," in Proc. 12th Annu. Rail-
road Engineering Conf., FRA/ORD 76-243, 1975.
spatial frequency, [23] D. Ahlbeck, "Preliminary evaluation of rail vehicles by computer
r time in which trucks pass a given point on the rail when simulation," J. High-Speed Ground Transport., vol. 11, p. 283,
moving with speed v, 1977.
[24] A. Prud'homme, "The track," French Railway Tech. Quart. Rev.,
r Llv, where L is distance between truck centers. no. 2, 1970.
[25] J. Corbin and W. Kaufman, "Classifying track by power spectral
The function Z(w) 12 is the track characteristic or square density," ASME AMD, vol. 15, 1975.
of the PSD.
ACKNOWLEDGMENT
Stanley A. Bernsteen received the B.S. degree
The authors acknowledge the consultation of the following from Cooper Union, New York, NY, in 1975, the
individuals who provided needed advice: Dr. Paul Christiano, M.S. degree from Carnegie-Mellon University,
Professor of Civil Engineering, Carnegie-Mellon University; Pittsburgh, PA, in 1977, both in physics, and the
Ph.D. degree in applied physics from Carnegie-
Dr. Don Ahlbeck, Battelle Columbus Laboratories, Track Mellon in 1979.
Characteristics and the Battelle Model; Dr. Shiv Gupta, Tim- S S Upon graduation, he joined Texas Instruments,
ken Bearing Company, Roller Bearings; Dr. A. Murphy, . where he works in systems analysis, mathemati-
cal modeling, and applications of artificial in-
Professor of Mechanical Engineering, Carnegie-Mellon Uni- telligence. He is currently seeking advanced
versity. degrees in computer science at the University of
Texas at Dallas.
REFERENCES
[1 Association of American Railroads, Yearbook of Railroad Facts,
1979.
[21 S. A. Bernsteen, "A theoretical analysis of rolling resistance of
Richard A. Uher (M'72), from 1968 to 1975,
railway vehicles from fundamental mechanics," Ph.D. thesis,
Carnegie-Mellon Univ., Pittsburgh, PA, Sept. 1979. was associated with the Transportation Division
131] W. J. Davis, "The tractive resistance of electric locomotives and of Westinghouse Electric Corporation where he
cars," Gen. Elec. Rev., 1926. held positions of Project Manager and Division
[4] N. Cooperrider, "Railway truck response to random rail irregular- Reliability Manager. Among the projects under
ities," ASME 74, WA/RT2, 1974. his supervision were propulsion equipment for
[5] D. Hannebrink, et al., "Influence of axle load, track gauge and _Sao Paulo, Brazil, subway propulsion equip-
wheel profile on rail-vehicle hunting." ment for New York R-44 cars, the Metroliner
[6] G. Doyle and R. Prause, "Hunting stability of rail vehicles with updating program, and advanced propulsion pro-
torsionally flexible wheelsets," ASME 75, WA/RT 2, 1975. jects. Since 1975 he has been associated with
[7] J. Hendrick, N. Cooperrider, and E. Law, "The application of Carnegie-Mellon University, Pittsburgh, PA,
quasi-linearization techniques to rail vehicle dynamic analysis," acting as principal investigator of several rail related federal contracts,
FRA/ORD 78-56, Nov. 1978. including energy management for electric powered transportation systems
[8] J. Siddell; M. Dokainish, and W. Elmargby, "On the effect of and improved passenger train service.
track irregularities on the dynamic response of railway vehicles," He is presently the Director of the Rail Systems Center and Senior
ASME 73, WA/RT 1, 1973. Lecturer, Physics and Electrical Engineering at Carnegie-Mellon Univer-
[9] "Dynalist II: A computer, program for stability and dynamic re- sity. His areas of expertise include railroad electrification, propulsion and
sponse analysis of rail vehicle systems," FRA/ORD 75-22-lV, July auxiliary equipment for main line railroads, rapid transit, light rail and
1976. people movers, energy management of railroads and transit systems and
[10] K. Lawson et al., "Dynamic railcar simulation program," FRA/ reliability and maintainability of transportation systems and equipment.
RT 70-24, Feb. 1970. He has authored over 35 papers and reports.
[11] A. Cohen and W. Hutchens, "Methods for the reconstruction of rail Mr. Uher is a member of the American Institute of Physics, the Trans-
geometry from mid-chord offset data," Trans. ASME, vol. 24, portation Research Board-Electrification Committee, and the American
1970. Public Transit Association-Rail Transit Committee.
[12] D. Ahlbeck, R. Prause, et al., "Comparative analysis of dynamics
of freight and passenger rail vehicles," FRA/ORD 74-39, Mar.
1974.
[13] D. Ahlbeck, Battelle Labs., Columbus, OH, personal communica-
tion. James P. Romualdi was Director of the Trans-
[14] D. Ahlbeck, H. Harrison, R. Prause, et al., "Evaluation of analyti- portation Research Institute at Carnegie-Mellon
cal and experimental methodologies for the characterization of University, Pittsburgh, PA. He is now Professor
wheel-rail loads," FRA/ORD 76-276, Nov. 1976. of Civil Engineering at Carnegie-Mellon and a
[15] S. Desai and T. Christian, Numerical Methods in Geotechnical Consulting Engineer. He has published in the
Engineering. New York: McGraw-Hill, 1977. area of structures, vibrations, fracture mechan-
[161 A. Jumikis, Theoretical Soil Mechanics. New York: Van ics, transportation, and engineering education.
Nostrand-Reinhold, 1969.
[17] T. Baumerster and L. Marks, Standard Handbook for Mechanical
Engineering. New York: McGraw-Hill, 1967.

Authorized licensed use limited to: Central Queensland University. Downloaded on July 25,2022 at 06:41:24 UTC from IEEE Xplore. Restrictions apply.

You might also like