You are on page 1of 32

Journal of Wind Engineering and Industrial Aerodynamics, 40 (1992) !

47-178 147
Elsevier

Aerodynamic drag reduction of goods trains

S. Watkins, J.W. Saunders and H. Kumar


Department of Mechanical Engineering, Royal Melbourne Institute of Technology, Melbourne,
Vic. 3001, Australia
(Received November 16, 1990; revised version accepted September 13, 1991 )

Summary
Tests on one-tenth scale hopper and gondola rail cars have been undertaken in a closed-circuit
wind tunnel with a fixed floor. By varying the number of wagons in the train length, it was con-
cluded that simulation of a typical wagon away from end effects (i.e. the influences of the loco-
motive (s) and trailing wagon ) could be achieved with one and a half "dummy" wagons upstream
and one half a "dummy" wagon downstream. An active model coupled to a drag balance was used
to measure drag coefficients and drag reductions from geometric modifications on a typical wagon.
The largest drag savings arose from fitting an enclosing top. Other effects, including inter-wagon
gap and end fairings were also investigated. It was found that a practical wagon design, that was
based on an existing wagon and allowed the current methods of loading and unloading, would
reduce averaged drag coefficients by 27% for an unladen wagon. Considerably larger savings (up
to 50% ) could be made with more substantial wagon redesign.

1. Introduction

Train aerodynamics have a long history; the first recorded wind-tunnel test
of a train appears to be in 1896 [ 1]. Since then, many workers have detailed
the results of wind-tunnel tests, although much of the recent work has concen-
trated on the drag and stability of high-speed passer~er transport. Recent wind-
tunnel studies on goods trains include: Hammi~t [2 ] who tested simple fiat
and block-shaped wagons and found large increases of drag coefficient with
yaw angle, Cataldi [3 ], who tested coal-carrying gondola and hopper wagons,
found that considerable drag savings could be made by covering the tops of
unladen wagons, and Engdahl [4 ], who tested open top hopper and gondola
cars with verification by full-scale tests, whose main findings included:
(i) inter-wagon gap and wagon length were important in determining drag,
(ii) drag increased substantially with reducing wagon load,
(iii) covering the tops of wagons gave an average drag reduction of 42%.
(iv) significant reduction in drag resulted from the addition of smooth sides.
Other studies performed with the support of Association of American Rail-

0167-6105/92/$05.00 © 1992 Elsevier Science Publishers B.V. All rights reserved.


148

roads, have investigated effects of the test-wagon position in train (see Gielow
et al., [5]).
The series of tunnel tests described here document the drag coefficient char-
acteristics of typical open-top gondola and hopper wagons including the effects
of wagon modifications. The wagons are as used by the various state rail sys-
tems in Australia to haul coal, wheat and iron ores in unit trains (i.e. trains
comprised of similar wagons). Although the results are specific to the wagons
modelled, it is considered that many of the results can be applied to general
wagon design.

2. Test methods

2.1. Wind tunnel


All the model-scale testing was carried out in the Industrial Wind Tunnel in
the Department of Manufacturing and Process Engineering at the Royal Mel-
bourne Institute of Technology (RMIT). The tunnel has a closed-return, a 2
m high by 3 m wide closed working section, and a porous roof (5% open area
ratio) r.nning the 9 m length of the test section.The testvelocitywas approx-
imately 30 m/s and the longitudinal rms turbulence intensitywas 1.5% in the
location of the model.

2.2. Models
Two one-tenth scale models were tested: a 16.871m long hopper wagon (des-
ignated Baseline 1 and based on the NHJF coal wagon as used by the state rail
system of New South Wales, see Fig. 1) and a 1.2.60 m long gondola wagon
(designated Baseline 2 and based on the P353 wheat wagon of the Queensland
Railways, see Fig. 2). The drag was measured on a modular, finely detailed,
"active" model with a series of dummy models either end of the active model
to simulate upstream and downstream flow conditions. Unless otherwise stated,
all models were tested unladen.

2.3. Track and force balance system


The models were mounted on a one-tenth scale track constructed on the
floor of the tunnel. Dummy models were fixed directly to the track (at appro-
priate wagon-to-wagon gaps) and the active model was mounted via a single-
component force balance onto the track. Small recesses in the track ensured
that the wheels of the active model did not make contact with the track and
hence interfere with the correct operation of the force balance. The entire sys-
tem was able to rotate (~o generate the desired yaw angle) about the centre of
the active model, thus keeping the centre of the active model in the same po-
sition in the airstream, irrespective of the yaw angle. The yawing system was

~AII dimensions are given in equivalent full-scale.


149

Fig. 1. Baseline I with half load.

iiii~i~!~i~
~ ¸¸

Fig. 2. Baseline 2 wagon ends.

driven by a lead screw, initially with manual operation and subsequently under
a computer-controlled stepping motor drive.
The single component drag balance was based on a pair of vertical strain-
150

gauged cantilevers operating in sway modes. Interactions from other forces


(e.g. lift and sideforce ) were found to be negligible and several calibrations and
baseline checks showed excellent repeatability and linearity. An IBM PC and
Data Translation DT2801A analog-to-digital card were used in conjunction
with Turbo Pascal programmes to measure and process the signals to drag
coefficient form. All drag coefficients were based on the unyawed baseline pro-
jected area.

2.4. Test procedure


The drag coefficient was measured at constant Reynolds number as the yaw
angle was varied. Initial tests included varying the Reynolds number and tur-
bulence intensity. Full-scale vehicles operate at Reynolds numbers approxi-
mately six times the model test Reynolds number. Whilst there are no general
guidelines for the testing of rail vehicles, the SAE (US) recommended proce-
dure for wind-tunnel testing trucks and buses indicates that 0.7X 106 is the
minimum acceptable test Reynolds number based on width, SAE [6 ]. To in-
vestigate the effect of Reynolds number further, the Baseline 1 model was tested
in the Reynolds number range (0.1-0.7) × 106 at various yaw angles. Whilst a
change in the drag coefficient occurred for a rounded top (not reported here),
the changes in Baseline 1 appeared minimal for ~ = 0 Aad ~ = 6 but the drag
coefficient clanged slightly at ~,= 15. It ~ppears that the Reynolds number
effects should be negligible. All subsequent testing was conducted at 0.7 × 106.
In full-scale, all vehicles will be subject to the turbulence in the natural wind
and the wakes of trackside objects. It was decided to investigate this effect,
since recent work comparing model and road drag reductions on commercial
vehicles (arising from aerodynamic drag saving devices) has shown this effect
to be significant, Watkins et al. [9 ]. A turbulence-generating grid was installed
well upstream of the models. The tunnel was recalibrated and the longitudinal
turbulence intensity measured at 3.7%. This intensity is consistent with typi-
cal train speeds and a moderate wind (although the turbulence scale will prob-
ably be incorrect). Baseline 1, and Baseline 1 with a flat top fitted, were tested
under these c~nditions. Comparing this with data taken with no grid in the
tunnel (as used for subsequent testing when a turbulence level of 1.5% was
present) showed slight changes in the shape of the graph but relative savings
were similar.
Testing also included visualising the flow around the baseline vehicles using
wool tufts and neutrally buoyant helium bubbles to highlight areas of separated
flow. Modifications to reduce the areas of separated flow were then tested.
The procedure for measuring and recording the drag coefficients was to align
the track with the mean flow direction (i.e. ¥ = 0 ) , set the tunnel speed at
approximately 30 m/s and sample data for 30 seconds at a frequency of 25 Hz.
Software processed the data to give the average drag coefficient. This process
was continued to give the drag coefficient from - 1 5 ° to + 15 ° in 3 ° incre-
ments. Any modifications were incorporated in each wagon.
151

3. Results

3.1. Data presentation


Drag coefficients for each modification are given as a function of yaw angle.
The general format for graphical presentation is as depicted in Fig. 3, the
straight lines joining the data are not intended to indicate any functional form.
All the rail wagons tested were nearly symmetrical and since the wagons can
travel in either direction, the data have been averaged for each positive and
negative yaw angle (i.e. data have been reflected about the vertical axis, then
averaged).

3.2. Position in train


Due to the large number of similar wagons in a unit train, the total train
aerodynamic drag can be estimated by the product of the drag of a "typical"
wagon (i.e. a wagon away from the influence of the front and rear of the train )
and the number of wagons in the train. Whilst individual wagon drag generally
reduces slightly with increasing distance down the train length, as shown by
Gielow et al. [5], the modelling of a full length train would produce such a
small scale model that the effects of Reynolds number would produce unreli-
able results and the ratio of model height to tunnel boundary layer thickness
would be unacceptably high.
For all tests, the drag coefficients were only measured on the active test
model. To investigate the effects of upstream and downstream flow modelling,
the dummy models were placed in various positions upstream and downstream
of the test model (Fig. 3). With no models upstream and half a model down-
stream, the drag coefficient was clearly at its largest, this would be considered
similar to the drag coefficient that a leading locomotive would experience, al-
though obviously the shape was incorrect. The next largest drag coefficient
was measured with 1.5 upstream models and no downstream models. This sim-
ulates the drag on the last wagon in a train. The two lowest drag coefficients,
with either one, or 1.5 dummy models upstream and 0.5 downstream are sim-
ilar, indicating that only minor changes should occur for wagons positioned
further along the train. It was decided that the most realistic simulation of a
wagon that would be well away from "end" effects would be to carry out all
subsequent testing with 1.5 models upstream and a 0.5 model downstream.
Further tests modifying the flow at the front of the leading dummy wagons had
negligible influence on the drag coefficient of the active model.

3.3. Wind-averaged drag


The method of wind-averaging the drag coefficient (WAD) has been used
to provide a long-term single drag coefficient that is relevant to given wind
conditions and vehicle speed, ~ee for example SAE [6 ] from Buckley et al. [ 7 ].
1.4 f~
b~

1.3-
~ ,,...~

1.2

1.1
/ NON :" UPS FRF___...AIv0.5 ;)OWN'. 3REAl

t-
Z
hi
0 0.9
h 1 s u P . ~ ~ ! NoN D O V , N S m ~ ~
I.d 0.8 .1/
0
0
0
,, "--,.
0.7
'"~-~-- '="- ------~i;~ ~ , u ~ .m~r ~::;; .,...~
0.6

0.5

0.4-
1 Ul 'STRE/ 0,! r DOW ~iS'TRE ~,M
0.3

TEST KX)G.
0,2 -

-16 - 12 -8 -4 0 4 8 12 16

YAW ANGLE (DEGREES)

Fig. 3. Effect of position in train.


153

Assumed conditions were:


(i) a vehicle speed of 70 km/h,
(ii) a wind speed of 11 kin/h, representative of the annual average wind
speed for USA or Australia, and
(iii) equal probability between wind and track direction.
An estimation of the angle subtended by the annual average wind and the
wagon speed can be calculated from SAE [6 ]. Since this value is typically about
6 °, some indication of the relative importance of the various modifications can
also be estimated from the graphs at ~ = 6 °.

3.4. Loads on baseline wagons


Generally, all tests were performed with the wagons unladen. To study the
aerodynamic effects of loads, two representative loads were te~ted in both base-
line vehicles. Figures 4 and 5 show the effects of loads on Baselines I and 2.
The highest point of the high load was 0.3 m above the top of the wagon and
its lowest level (where the top surface of the load met the wagon sides ) was 0.3
m below the top of the wagon. The intermediate load was the same shape as
the high load, but the top of the load was just level with the top of the wagon.
To compare the difference between wagon types, Fig. 6 shows data for both
baselines. Clearly, Baseline 2 (the gondola) has a much lower aerodynamic
drag than Baseline 1 (the hopper ), whether laden or unladen. The wind-av-
eraged drag values were 0.474 for Baseline 1 empty and 0.330 for Baseline 1
with full load. Baseline 2 values were 0.277 and 0.171 for unladen and laden
wagons, respectively.

3.5. Inter-wagon gap


The effect of the gap between wagons is given in Figs. 7 and 8 for Baseline 1
and 2. The gap spacing was defined as the distance from the end of the wagon
chassis to the start of the adjoining wagon chassis.
To clarify the effects, the data for Baselines 1 and 2 have been crossplotted
for yaw angles of 0 °, 6 ° and 12 °, see Figs. 9 and 10. Baseline 1 and 2 wagon
gaps of 0.83 m and 0.65 m were used for all subsequent testing, which was
consistent with current Australian usage.

3.6. Wagon length


The length of the Baseline I wagon was reduced from the standard baseline
length of 16.87 m to 11 m. Reducing the length reduced the drag coefficient
significantly {Fig. 11), particularly at high yaw angles, but obviously reduces
carrying capacity, dictating a greater number of wagons to haul any given load.
The wind-averaged drag value for the shortened wagon was 0.383.
0.75 ¢Jl
I I I 1 I " I I I |

0.7 /
r-- ,-'1

0.65
\ /
0.6
BASELINE1(NHJF) 2'
WITHHIGHANDLOWLOAD i/
\
i 0.55
~ A ' ~ ~ ~'~B A B '
1~ HiGHI 0.30 0.30I(M)
/
LOW l o eo ooo I / /
O 0.5
O //
,,"t
~; 0.45
/
\
0.4
/j /
0.35
//
0,3 ' .oot J
0.25 I I !
,==='
-8 -4 0 4. 8 12 16

Y A W ANGLE (DEGREES)

Fig. 4. Effect of loads on Baseline 1.


0.4.4
J i I I I I I I I
0.42 u

_ .,//
0.4
/,"
0.38

0.36
\ /
BASELINE 2 (P353 GONDOLA)
0.34 ....~ ?
0.32
,, \ ^-f-" "-~F I /
bJ f
N
u +,, / ),.
r;" 0.3 \ .,o.10.~0 /
i
It. ~ cowl0..00.0
0.=1<.1 / i
Id 0.28
\ \ / /
ou /
0.26 -\ ' \ /
/ /)"
0.24.
\ \\ /
0.22 \

0.2
\ \ /
/
0.18

0.16
j ./
0.14
0.12
-16 -12 -8 -4 0 4 8 12 16

YAW ANGLE (DEGREES)

Fig. 5. Effect of loads on Baseline 2.

p.d
0.8" fj1
I o~
I
I
o i
/
0.7
~- -' t
BASELINE I (NHJF)

0.6

! ! ,

0.5

o 0.4

£:3

0.3

E 2 (P~.~ GONOOI..A)
0"2I LNI2T'~ I >. ~
I~
,A. . . . .,,,.,,~,,.~,.i . . ~ | ., ~
0.1 t I l I I 1 l i ,
i i ! i ]
-16 -12 -B -4 0 # 8 12 16

YAW ANGLE (DEGREES)

Fig. 6. B a s e l i n e I a n d 2 - Comparison.
o.g

o.8 ~,P-l.~7m , ! /,,~


~ 1 ~ 1"27n~ /

CJ
g~

0
cJ

D
L
C3
0.5

0o~ "

0,3 "
I
-16 - 2 -8 -4 0 4 8 12 16

Yaw Angle (Degrees)

Fig. 7. Effect of gap on Baseline 1.

....1
0.55 t==i
i ! I I O1
I I t ] ....~ oo

r-
0.5 2
L._ ]

0.45

0.4
\,
kml
0 ~k, EFFECTI]/EGAP
00 0.35
/
0.3
. J'
i
~'~ 1
0.25

0,2 " ,

0.15 !6
! !
-8 -4 0 4 8 12 16

YAWANGLE(DEGREES)
Fig. 8. Effect of gap on Baseline 2.
159

P4
I i

B
00

il
IF W
._I

Cn
- .._I
•- .....j

I,

r~
C I,i
l--
C~l W

>

- 0

Z
0

L
J
I

I
_ o
I-.
I
- Ip Z
QO 0
d o !

~0
0

.l,.q

0
0
0 6 o o ci o o o

.l,,~

.LN31O1-1_-1300 OV~ICI r,z-,


0.9

0.8

0.7
BASELINE 2 (P353 GONDOLA} WITH TOP
I-
Z
w 0.6
C.)
b_
b_
W 0.5
O
L) YAW = 1 2 ~ . - - . - ' e - ~ ~ ' - "
0
< 0.4 C~
0
YAW_.~~ ......---F- ~
0.3
. YAW~._.D..--- ~ ~ " ~ "
0.2

0.1

0.6 0.8 1 1.2 1.4 16 18

WAGON TO WAGON GAP (METERS FULL SCALE)

Fig. 10. Drag coefficient versus wagon-to-wagon gap - Baseline 2.


DRAG COEFFICIENT
0 0 O 0 O 0 0
I
--.A
o
I
I
I-"
rrl
~3
m 0
N
00
\
191
162

3. 7. Top coverings
Considerable savings can be achieved with simple coverings over the tops of
open, unladen vehicles (e.g. see Cataldi [ 3 ] and Watkins et al. [8 ] ). Initially,
it was decided to test only plane tops; however, since the drag reductions were
found to be considerable, further top modifications were investigated to take
the loading of wagons into account. Many wagons are loaded automatically,
via a single chute which frequently is significantly narrower than the wagon
width. Hence, the opportunity arises to partially cover wagon tops to reduce
drag, whilst still permitting practical automatic loading.

3. 7.1. Partial fiat tops


Longitudinal gaps running down the centreline of the wagon (Fig. 12) of
various widths were tried, ranging from fully covered to fully open. Results for
fiat, non-porous coverings over the top of Baseline 1 are given in Figs. 13 and
14. As expected, the drag increases with increasing gap width. The correspond-
ing wind-average drag figures were 0.325, 0.392 ~nd 0.446 for no gap, 0.6 m gap
and 1.70 m gap, respectively. The effect of completely covering the top of the
Baseline 2 wagon can be seen in Fig. 15.
The effect of removing the longitudinal sills along the sides of the wagons
was to further reduce the drag coefficient to a value of 0.1 at zero yaw angle.

3. 7.2. Partial angled tops


To enable the load carrying capacity to be enhanced, top coverings with
angled sides were fitted to the Baseline I wagons. The width o~ the gap in all

Fig. 12. Drag coefficient versus patti; ~fiat-top gap - Baseline 1.


0.8

%
0.7
le.87M '

1.70M GAP
,/
0.6
/
BASELINE I WITH FLAT COVERS

0.5

0.4
~ ",~---~---- ,

0.3 _ _ "~ , . _ _ 2 ' , ° , , _ ~,7


4L

Fop C )vere(

0.2
--16 --12 -8 -4 0 4 8 16

YAW ANGLE (DEGREES)

Fig. 13. Baseline 2 with 0.60 m slots.

i-,I
Cb
C~
0.8

0.7
I'AW= 1 ~!
......~ ..-
J
I-- 0.6
Z I
td
(3
b.
b.
W 0.5 J
f
0
U

YAW=6
~ .~...i....._ ....-..---~ ~ - - "
0.4.
YAW=C
I't

... j . ~ - - - - - ~ - --'---
i J
0.3 J

0.2
0 0.4. 0.B 1.2 1.6 2.4 2.9
SLOT GAP ( m )

Fig. 14. P a r t i a l fiat t o p s - Baseline 1.


0.44
0.42
0.4 -\
/ 1
J
0.38 .. l (.,.,j,_~v; /
0.36
\ - !
_/
BASELINE 2 (i>353 GONDOLA) WITH TOP
0.34
\ /
0.32
\ I
i,i 0.3 h__ I I -2'
(J % /
,n- 0.28
b. /
bJ 0.2.6
/
o
rj /
0.24 %~ \I l ,
o / /d/
/
0.22
a .X (jr
0.2 \ ~ F-~ p~- .¢"/
0.18
-~~ /'J
0.16
0.14
0.12
0.1 13
I
-I- TOF LES.' SIDE SILLS I
0.08
-16 -12 -8 -4 0 4 8 12 16

YAW ANGLE (DEGREES)

Fig. 15. Effect of tops on Baseline 2.

Ob
C,rl
0.75 - w i ! I ! I |

0.7

600mm GAP /
0.65

\ BASELINE 1 WITH ANGLED TOPS


/
0.6 600 mm GAP

I-
Z .~] TOP lil
w
m ANGLE
0.55
0
b_
b_
W
0
0,5
/

L9
< 0.45 ]
I...3
BASELINE I
0.4 ,q V
0.55 7 5 DEG

0.5

0.25 I
- 6 -12 -8 -4 0 4 8 12
YAW ANGLE (DEC)

Fig. 16. Effect of angled tops on baseline I - 600 m m slot gap.


167

these tests was fixed at 600 mm and the angle between the top surface and the
horizontal was varied, see Fig. 16. The wind-averaged drag figures were 0.371,
0.357 and 0.374 for angles 30 °, 37.5 ° and 45 o respectively. The optimum angle
of 37.5 ° was chosen for further work which wa.~ undertaken on Baseline 1 only.

3.7.3. Effect of gap width on 3Z5 ° angled top


Figure 17 shows that width increasing gap width, the drag coefficient also
increases and, except for 600 mm gap configuration, the drag coefficient did
not vary significantly at zero yaw angle. However, above 3 ° the savings are
substantial.

3. 7.4. Effect of angled tops with 1350 mm gap


Many loading situations dictate a wider gap and on advice from the State
Rail Authority of NSW, a 1350 mm gap was chosen as a practical width. The
two angle configurations tested were 37.5 o and 45 ° tops. Figure 18 shows that
the 37.5 ° top was again better at all yaw angles and the wind-averaged drag
values were 0.406 for 37.5 ° angled top and 0.429 for 45 ° top.

3.8. End modification


A radiussed fairing, comprising of a section of a cone, was fitted to the wa-
gons with 37.5 ° angled tops and 1350 mm gap, see Fig. 19. As can be seen in
Fig. 20, the drag coefficient is reduced at all yaw angles including zero. The
wind-averaged drag was reduced from 0.406, with angled tops only, to 0.346
with the tops and ends. No other shape of end was considered since other end
geometries, such as compound curvature fairings, have been investigated in
prior work (Watkins et al. [8]). A three-dimensional (double-compound)
curved surface may produce slightly better performance, but it would be ex-
pensive to produce, whereas the conical fairing can be conveniently rolled from
sheet metal. This configuration was designated the optimum wagon.

3.9. Loads on optimum wagon


The effect of loads on this optimum wagon was relatively small compared
with the effect on an empty baseline wagon, see Figure 21. The wind-averaged
drag values for the partially-laden and fully-laden wagons were 0.357 and 0.325.

3.10. Optimum modifications on shortened vehicle


The effect of 37.5 ° tnp with 1350 mm gap on the shorter (11 m) baseline
was similar to the savings found on the longer ( 16.87 m ) baseline. The effect
of ends had a similar trend. Figure 22 shows the shape of the curves and, as
can be seen, the radiussed ends reduced the drag coefficient values substan-
tially at all angles.
U.O

oo

~1 I GAP WIDTH /
....
0.7 --
! L
~~% BASELINE1 WITH 37.5 DEGREEANGLEDTOP

I- '~ 37.5o
Z
w
w 0.6
O
b_
b_
W
0
O

0.5
/
n-
O
I
180OHM "~

0.4
;Ylli I
, 1350MM

0.3 "

-16 -'2 -8 -4 0 4 8 12 16

YAW ANGLE (DE(3)

Fig. 17. Effect of gap width on angled tops.


0.75 I i I i I I I i I

.7
[~. • ~ 13S0mm GAP --..~

0.65
BASELINE 1 WITH ANGLED TOPS
1350 mm GAP
I
I-.
0.6
Z TOP
Ld ANGLE
0
0.55 -
t'-
Ld
0
0
0.5 \
0

0 B A S E L INE I .-_=
0.45

0.4
37.5 D E G

0.35

DEG
0.5 I
-16 -8 -4 0 4 8 12 16

YAW ANGLE (DEG)

Fig. 18. Effect of angled tops on Baseline 1 - 1350 m m slot gap.

t-,a,
¢,,o
170

Fig. 19. Baseline 1 with optimum configurations.

3.11. Deflectors and fairings


Baseline 1 was fitted with two different end add-oas, to try to reduce the
amount of air entering the unladen wagon. A quadraTlt section (of radius 0.60
m in side elevation ~ spanning the width of the wagon, was attached on the
upstream side only, as shown in Fig. 23. Whilst it ~:¢asrealised that this drag-.
reducing device would only reduce the drag when the vehicle was travelling in
one direction (it increases the aerodynamic drag when travelling in the oppo-
site direction), some wagons are unidirectional.
A separate test was performed to assess ÷¢heeffect of fairings fitted into the
ends of the wagons. Both ends of the wagon were faired with streamlilie sec-
tions of length 0.60 m and depth 0.3~ m.
The deflector made negligibl.c change at yaw angles of up to six ° but showed
~light savings at nine d.cgrees and over. The fairings gave a slight drag reduc-
tion of approximately 0.025 throughout the yaw angle range tested.
! !
rO.lb I I I I I I

0.7

0.65

BASELINE 1 WITH 37.5 DEGREEANGLEDTOPS


0.6 1350 mm GAP & FRUSTRUM ENDS

0.55 1350 mm GAP


t
0.5

0.45

BASELINE 1
0.4

0.35 . . . . . . . . . . ANGLED TOP5


0.3

0.25
o~ ~ ~o~
1

- 6 -12 -8 -4 0 4 8 12 16

Y A W ANGLE (DEG)

Fig. 20. Effect of tops and ends.

--,.1
0.8
- - " J~__ FU~,'I L O A D HALF LOAD

% ~,~ .~03 M

0.7 ~ ~I.0M ,/1


E r-- T "-=1" -I
~x__/x__/Z_/~_/~
BASELINE 1 WITH 37.5 DEGREE ANGLED TOPS
0.6 I 1350 mm GAP & FRUSTRUM ENDS
L-.
m
Z
Lu
O
~1350 mm GAP
b_
Lu 0.5
0
(3
©
<
¢r BASEL I NE
0.4
I
I I1
I
/
HALF LOA[
1
0.3
I
NO LOAD

L LOAD
1
0.2 I I I I I I I ,1
-16 -12 -8 -4 0 4 8 12 16

YAW ANGLE (DEG)

Fig. 21. Effect of loads on optimum wagon.


0.58
I I I I I I I i I
0.56 / :. . . . \
0.54
0.52 C.'-- ~ "7
0.5
0.48
IX \ ~~ ~~-
Z
0.46
w Xl, I \. ~\ I I/r ' -'~
o 0.44
b_
b_ 0.42 -
\ I c ,,~,oooo.. 3~
W
0
o 0.4
0< 0.38
,'7"
0.36 t \k,,~--~l I " "
0.34
0.32
I -"~ I " - I ' I
0.3 I - . ~ ~''" ANGLED TO~S

0.28

0.26
0.24 ,1I .... I
-16 - 2 -8 -4 0 4 8 12
YAW ANGLE (DEG)

Fig. 22. Effect of tops and ends - Shortenedwagon.


zf "~ -O

0.8 -1
T.... F- I "I i l---l- I _l~-
BASELINE I WITH DEFLECTOR
I

0.7

~ 0.6mj
0.6 -, 1
I f-
I I ,
! ---1 I-
- - - r - - ~
-I
-
i
0.3m

0.5 . . . .
I I
0.4 -, , WITH FAIRING

I _
0.3

0.2 , I i ' J--t--, ,H--~


-16 -12 -8 --4 0 4 8 12 16

YAW ANGLE (DEGREES)

Fig. 23. Deflector and fairings.


175

4. Discussion

4.1. Drag reductions


In order to clarify the average drag savings that might be expected from the
tunnel data, Table 1 presents the wind-averaged drag coefficients.
Clearly, the baseline drag coefficient of the gondola wagon was much lower
than the hopper, irrespective of load. This was considered to be due to a com-
bination of the wagon's shorter length, different end geometry and more
streamlined underbody loading geometry.
For any wagon, large drag reductions are evident from enclosing the open
top, either by a load, or a partial or complete top; deflectors fairings such as
the conical segments fitted to the wagon ends yielded significant drag reduc-
tion, whereas smaller, end covering fairings and deflectors fitted over the load
area showed only minor savings. Reducing wagon-to-wagon gaps also showed
benefits in drag savings, as well as potentially reducing loading and unloading
times. Reducing wagon length obviously reduces drag but also reduces load,
hence an economic decision needs to be made based on specific operations.
The optimum practical wagon gives significant wind-averaged drag savings of
27%, compared to an empty baseline, whilst changes in railway operations
could allow practical enclosed wagons of drag coefficients approaching 0.1.

4.2. Wind-tunnel simulation


A fixed ground simulation was used, resulting in a boundary layer height
that was of the order of a wheel diameter. This would serve to slightly under-
estimate the total drag coefficients, since in computing the drag coefficient, a
flat profile was assumed ( i.e. the boundary layer velocity deficit was ignored).
However, all the modifications were well away from this area, hence it is con-
sidered that the absolute drag saving would be negligibly affected, whereas the
percentage savings may be slightly overestimated due to a baseline drag coef-
ficient that was slightly too low. Blockage corrections have not been applied to
any of the data, since the blockage ratio was low (model area was less than 2%
of test area at zero yaw ) and the roof of the tunnel was partially vented. A
longitudinal static pressure gradient coefficient of - 0.0078 m - 1was measured
at the model position, which would tend to increase the measured drag coeffi-
cient on all configurations by typically 0.010. This correction has not been
incorporated.

4.3. Implications for fuel saving and wagon design


For long trains, the total train drag can be estimated by multiplying the
individual wagon drag (as measured here) by the number ~f wagons ~in the
train, if boundary layer growth along the train has a negligible effect. This
seems reasonable for bluff wagons where pressure drag is the dominant drag
term and this was found to be the case by Gielow et al. [5] at zero yaw. An
176

TABLE 1

Wind-averaged drag coefficients


m

Baseline 1 configuration Wind-average % reduction


drag from baseline

Wagon to wagon gap 0.83 m - (Baseline 1 ) 0.474


- low load 0.374 21.1
- high load 0.330 30.4
Wagon to wagon gap 0.55 m 0.472 0.4
Wagon to wagon gap 1.27 m 0.494 -4.2
Wagon to wagon gap 1.87 m 0.521 -9.9

Flat tops - no gap 0.325 31.4


Flat tops - 0.60 m gap 0.392 17.3
Flat tops - 1.70 m gap 0.446 5.9

Fairings both ends 0.447 5.7


Deflector one end 0.465 1.9

30 ° angled top - 0.60 m gap 0.371 21.7


45 ° angled top - 0.60 m gap 0.374 21.1
37.5 ° angled top - 0.60 m gap 0.357 24.7
37.5 ° angled top - 0.90 m gap 0.384 19.0
37.5 ° angled top - 1.80 m gap 0.435 8.2
37.5 o angled top - 1.35 m gap 0.406 14.3
37,5 ° angled top - 1.35 m gap ends 0.346 27.0
- half load 0.357 24.7
- full load 0.325 31.4
Shortened wagon len~,,h 11 m 0.385
37.5 ° angled top - 1,35 m gap 0.346 10.1
:37,5 ° angled top - 1,35 m gap ends 0.298 22.6

Baseline 2 configuration Wing-average % reduction


drag from baseline

W a g o n to wagon gap 0.65 m - (Baseline 2) 0.277


- low load 0.209 24.5
- high load 0.171 38.3
W a g o n to wagon gap 0.87 m 0.302 -9.0
W a g o n to wagon gap 1.27 m 0.341 -23.1
W a g o n to wagon gap 1.87 m 0.408 -47.3
Flat tops 0.163 41.2
Flat tops - no side sills 0.142 48.7

Note: Unless stated otherwise all wagon configurations are unladen.

allowance can easily be made to incorporate wagons or locomotives that are in


the influence of the ends of the train.
Savings arising from the modification of the vehicle tops are generally low
177

for a fully laden wagon, whereas the drag savings on an unladen wagon are
substantial. Since most rail vehicles travel half the time unladen, the overall
savings would be approximately half the unladen wagon savings.
At typical train speeds, the fuel savings on a level track at constant speed
should be slightly over half the drag savings, since under level, steady condi-
tions, Joshi [ 10 ], found that between 50% and 70% of power is used to over-
come aerodynamic drag. Clearly, detailed predictions will depend upon the
specific operating environment. A mathematical model has been developed to
assess fuel savings on Australian routes.
The optimum wagon design offersadvantages of improved load capacity, as
well as minimising dust into and out of the wagons. Patents are pending for
partialtop configurations in Australia,Brazil,Canada, India, South Africa and
the USA.

5. Concluding remarks

Current goods wagon design appears to be dictated by loading, load max-


imising and structural considerations with generally littlethought for aero-
dynamic drag minimisation. The drag reduction methods detailed here show
significantsavings which are easilyincorporated into new wagons.
It is interesting to note that the potential for drag reduction on trains is
considerably greater than for road vehicles due to the higher base pressures
afforded by close-coupled vehicles.Additionally,rollingresistancecoefficients
for rail vehicles are substantially lower than for commercial road vehicles.
Quoted values of rolling resistance for rail vehicles vary between 0.0008 to
0.002 for speeds of 100 km/h [11 ] whereas corresponding values for commer-
cialvehicles fittedwith good radialply tyres vary between 0.006 to 0.003 [12 ].
Total energy usage per unit volume transported is thus much lower for trains
than for road haulage.

Acknowledgements

The authors would like to thank Mr. Michael O'Rourke, formerly of Rail-
ways of Australia Committee for his considerable assistance, Mr. John Mep-
stead and Mr. Kevin Gibson from R M I T for technical help and the many peo-
ple from all State Rail Systems in Australia for their contributions to this
work.
The financialassistance of the National Energy Research, Development and
Demonstration Program is gratefullyacknowledged.
178

References

1 W.F.M. Gross, Atmosphere resistance to the motion of railway trains, The Engineer (1898).
2 A.G. Hammitt, Aerodynamic Forces on Freight Trains, Volume 1. Wind Tunnel Tests of
Containers and Trailers on Fla:~cars, U.S. Department of Commerce National Technical In-
formation Series PB-264, 304, December 1976.
3 R.G. Cataldi, Aerodynamic design considerations for bulk trains, Proc. 3rd International
Heavy Haul Railway Conference, Vancouver, Canada, October, 1986.
4 R.A. Engdahl, Full-scale rail car testing to determine the effect of position-in-train on aero-
dynamic resistance. Report No. R-705, Association of American Railroads Research and Test
Department, November, 1987.
5 M.A. Gielow and C.F. Furlong, Results of wind tunnel and full-scale tests conducted from
1983 to 1987 in support of the Association of American Railroad's Train Energy Program,
Report Number R-685, Association of American Railroads Research and Test Department,
December, 1988.
6 Anon, SAE wind tunnel test procedure for trucks and buses, SAE (US) J1252, published by
the Society of Automotive Engineers, USA, 1979.
7 F.T. Buckley, C.H. Marks and W.H. Walston, A study of methods for improving truck fuel
economy, prepared for the National Science Foundation under Grant SIA-74-14843, Uni-
versity of Maryland, USA, 1978.
8 S. Watkins and J.W. Saunders, Aerodynamic resistance and design principles for unit train
rail vehicles, Department of Manufacturing and Process Engineering, Royal Melbourne In-
stitute of Technology, 1988 ( ROA Rep. No. 14/ 1/ 88 ).
9 S. Watkins, Wind tunnel and road tests on add-on aerodynam~ devices for commercial ve-
hicles, WTP 1041, Department of Mechanical and Production Engineering, Royal Mel-
bourne Institute of Technology, 1986.
10 P.B. Joshi, Aerodynamic Forces on Freight trains, Volume II, Full-Scale Aerodynamic Vali-
dation Tests of Trailer-on-a-Flat Car, U.S. DOT-FR-78-19, 1978. Available from the U.S.
National Technical Information Service, Springfield, VA 22161, USA.
11 G. English, J. Young, H. Boumeester, C. Schwier, M. Roney and P. Bunting, Railway linehaul
energy intensity: An analysis leading to design of a train simulation software package, Ca-
nadian Institute of Guided Ground Transport Rep. 80-15, 1981.
12 J. Ramshaw and T. Williams, The rolling resistance of commercial vehicle tyres, Transport
and Road Research Laboratory (UK) Supplementary Rep. 701, 1981.

You might also like