Wear 523 (2023) 204727

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

Slip of wheels on rails: The root cause for rail undulant wear
Guiming Mei a, Guangxiong Chen b, *
a
State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu, 610031, China
b
School of Mechanical Engineering, Southwest Jiaotong University, Chengdu, 610031, China

A R T I C L E I N F O A B S T R A C T

Handling Editor: Dr. M Dienwiebel In this paper, a comprehensive field investigation was carried out into rail corrugations of China’s metro lines
and railway main-lines. The occurrence rate of rail corrugation, which is defined as the accumulated length of
Keywords: corrugated rails divided by the total length of the investigated line, is about 11.94% for metro lines, or 2.73% for
Undulant wear railway main-lines. In terms of the viewpoint of the friction-induced vibration inducing rail corrugation, the
Rail corrugation
authors analyzed the reason to cause the difference in the occurrence rate of rail corrugation between the metro
Slip
lines and railway main-lines. The slip of wheels on rails is proposed as the root cause for rail corrugation. Without
Wheel/rail interaction
Friction-induced vibration the wheel/rail slip, no corrugation occurs. Model predictions of rail corrugation were performed on the sharply
curved tracks, mildly curved tracks and tangential tracks. Prediction results validated the conclusion. Finally, a
fast prediction method for rail corrugation was proposed.

1. Introduction curved tracks is close to 100%, whose radii are less than or equal to 350
m in China or elsewhere. Except for some passive methods of sup­
Early in 1920s, the rail corrugation was researched. In 1990s–2010s, pressing rail corrugation such as rail grinding, friction modifier and
the rail corrugation was investigated throughout the world. In this variation of passing speeds, no active method is exploited successfully.
period of time, the modern theory of rail corrugation was established, The authors visited many metro lines throughout the world and have
which mainly consists of the wavelength fixing mechanism and the found that many subway lines are plagued by the corrugation problems.
material removal mechanism. Many researchers have made big contri­ In China, almost all metro lines are troubled by corrugation problem. As
butions to the establishment of the rail corrugation theory [1–20]. long as the radius of curved tracks is less than or equal to 350 m, almost
Especially, Grassie has made great contributions [1,2]. Grassie and all low rails of the curved tracks are subjected to rail corrugation.
Kalousek proposed that excitation of the vehicle’s unsprung mass on the Therefore, more endeavors need making to solve the rail corrugation
track stiffness is the most common wavelength-fixing mechanism [1]. problem.
More than two-thirds of all publications on rail corrugation mainly In the paper, the authors made a brief review of the problems and the
involved this excitation from different perspectives. Nowadays, almost research theory of rail corrugation. Secondly, the authors presented field
all research works are carried out based on the modern rail corrugation investigations into the rail corrugation in the China’s railway tracks,
theory [21–34]. According to Grassie [1,2], the mechanisms for rail including railway main-lines and metro lines. The authors also analyzed
corrugation generation have been well understood. At present, the these field investigation results into rail corrugation. Thirdly, the au­
railway technology has been developed greatly. The running speed of thors proposed the slip of wheels on rails as the root cause for rail
passenger trains has reached 350–400 km/h. The axle load of freight corrugation, and performed several model predictions to verify the
trains has reached 25,000–30,000 kg. Every 10–20 years, the railway conclusion. Then, the authors proposed a fast prediction method for rail
technology has reached a new level. During the period time of 10–20 corrugation. Finally, the authors made several conclusions.
years, many old problems were solved, and many new techniques were
applied successfully. In the eyes of the authors, rail corrugation is the
most difficult problem in the field of railway technology. Although some
corrugation problems on mildly curved tracks and tangential tracks have
been solved, up to now the occurrence rate of rail corrugation on tightly

* Corresponding author.
E-mail address: chen_guangx@163.com (G. Chen).

https://doi.org/10.1016/j.wear.2023.204727
Received 18 September 2022; Received in revised form 16 November 2022; Accepted 6 December 2022
Available online 5 March 2023
0043-1648/© 2023 Elsevier B.V. All rights reserved.
G. Mei and G. Chen Wear 523 (2023) 204727

2. Field investigation into rail corrugation 2200–2400 mm. The wheel diameter of metro trains is 840–880 mm.
Several types of sleepers such as common short sleepers, booted
2.1. Field investigation into rail corrugation in China railway main-lines sleepers, and single-block sleepers are applied in metro tracks. Several
types of fasteners such as DTIII-2 fasteners, DTIV-2 fasteners, WJ-3
Rails of 60 kg/m are laid in China’s railway main-lines. The distance fasteners, Cologne-egg fasteners, and Vanguard fasteners are applied
between two concrete sleepers is 525–635 mm. The rail gauge is listed in in metro tracks. These fasteners provide different spring stiffness and
Table 1. Ballast tracks account for most of the China’s railway main- damping values for metro tracks. The vertical and lateral spring stiffness
lines, and high-speed tracks are equipped with monolithic beds, as values of Cologne-egg and Vanguard fasteners are about 8–12 MN/m
shown in Fig. 1. Since 2001, the minimum radius of curved tracks has and about 6–8 MN/m, respectively. The vertical and lateral spring
been changed to 800–1000 m from 300–500 m to run freight trains of stiffness values of DTIII-2 fasteners, DTIV-2 fasteners, WJ-3 fasteners are
speed 120 km/h. Generally, passenger trains and freight trains all run on about 20–40 MN/m and about 12–20 MN/m, respectively. Later on, the
the ballast tracks. Only high-speed trains run on the monolithic bed authors will illuminate that rail corrugation more easily occurs on the
tracks. The wheels of all passenger trains and freight trains are equipped track equipped with Cologne-egg fasteners based on the field investi­
with the worn-type tread profiles. The conical tread profile has been gation data.
abolished. The distance between two axles of freight car bogies is Rail corrugation more easily takes place on metro lines. Since some
1750–1830 mm. The distance between two axles of main-line passenger radii of metro curved tracks are inevitably less than 300–350 m, rail
vehicle bogies is 2400–2700 mm. The nominal wheel diameter of freight corrugation is most likely to occur on the rails whose radii are less than
trains is 840 mm. The nominal wheel diameter of passenger trains is 915 or equal to 350 m. According to the authors’ field investigation into rail
mm. The nominal wheel diameter of high-speed passenger trains is corrugation in China’s many metro lines, the authors found that almost
860–920 mm. The China’s main-line tracks are mainly equipped with all low rails suffer from rail corrugation on the curved tracks whose radii
No. 12-type turnout. The radius of the leading curve of the turnout is are equal to or less than 350 m, as depicted in Figs. 3–11, but only a
180–200 m. small number of high rails of the curved tracks with low rail corrugation
By the end of 2021, the total mileage of China’s railway main-line suffer from rail corrugation. Some low rails probably suffer from rail
tracks is about 150,000 km, which includes high-speed train tracks of corrugation on the curved tracks whose radii are equal to 450–1100 m or
40,000 km. The China’s railway main-line tracks of 150,000 km are on tangential tracks, as depicted in Figs. 12–14. When the radii of curved
administrated by 18 regional railway group corporations. Each regional tracks are larger than 650–800 m including tangential tracks, two rails
railway group corporation averagely administrates railway main-line of the tracks are seldom subjected to rail corrugation, as depicted in
tracks of 8,333 km. The authors investigated into rail corrugation in Figs. 15–17. From Figs. 3–17, it is observed that the tracks with Cologne-
two larger railway group corporations and obtain two representative egg fasteners most easily suffer from rail corrugation, in comparison
data. One railway group corporation administrates railway main-line with those tracks with other fasteners. Not only tightly curved tracks,
tracks of 10,257 km. The accumulated length of the corrugated tracks but also tangential tracks with Cologne-egg fasteners are subjected to
is about 280 km. Most of the rail corrugation occurs on the leading rail corrugation. Not only low rails, but also high rails are subjected to
curves of turnouts, whose radius is 180–200 m. The occurrence rate of rail corrugation.
rail corrugation for railway main-line tracks is equal to about 280/ The authors made several field investigations into rail corrugation
10257 = 2.73%. Since the occurrence rate of rail corrugation on the and took many photographs as shown in Figs. 3–17. The authors’ field
railway main-line tracks is so small, one can conclude that rails of rail­ investigation into rail corrugation has proved to be a very hard work.
way main-line tracks are not easy to suffer from rail corrugation. Fig. 2 That is why few publications on field investigations into metro rail
depicts some rail corrugations occurring on the China’s railway main- corrugation were reported in the literature. Zhang et al. [35] reported a
line tracks. Another regional railway group corporation administrates comprehensive field investigation into rail corrugation in Beijing metro
railway main-line tracks of 9,035 km. The accumulated length of the line 5. According to Zhang et al.’s field investigation data, the total
corrugated tracks is about 107 km. The occurrence rate of rail corru­ milage of Beijing metro line 5 is 53.69 km, among which the accumu­
gation for railway main-line tracks is equal to 107/9035 = 1.18%. lated mileage of tangential tracks is 37.21 km, the accumulated mileage
According to the statistical data of rail corrugation in two regional of sharply curved tracks whose radii are less than or equal to 400 m is
railway group corporations, it can be concluded that the occurrence rate 4.18 km, the accumulated mileage of mildly curved tracks whose radii
of rail corrugation on railway main-line tracks is about 1.18–2.73%. It are larger than 400 m is 12.3 km. Several types of fasteners were used,
can be seen that rail corrugation is not easy to occur on railway main- which included Cologne-egg and DVTI-2 fasteners. The Cologne-egg
line tracks. fastener was characterized by a low vertical stiffness, about 8–12
MN/m. The vertical stiffness values of other fasteners were 20–48
MN/m. The accumulated mileage of the track with Cologne-egg fas­
2.2. Field investigation into rail corrugation in China metro lines
teners was 20.55 km. The accumulated mileage of the track with other
fasteners was 33.14 km. In Beijing metro line 5, the accumulated
Rails of 60 kg/m are laid in China’s metro lines. The distance be­
mileage of corrugated rails was 6.41 km, in which the accumulated
tween two concrete sleepers is 580–650 mm. The rail gauge is listed in
mileage of corrugated rails with Cologne-egg fasteners was 5.16 km, the
Table 2. All metro tracks are equipped with monolithic beds. The min­
accumulated mileage of corrugated rails with other fasteners was 1.25
imum radius of curved tracks is larger than 100 m. The maximum su­
km.
perelevation is 120 mm. Generally, metro trains are controlled by ATO
The occurrence rate of rail corrugation in Beijing metro line 5 was
(Automatic Train Operation). The wheels of all metro trains are equip­
equal to 6.41/53.69 = 11.94%. On all tracks with Cologne-egg fasteners,
ped with the worn-type tread profiles. The conical tread profile has been
the occurrence rate of rail corrugation was equal to 5.16/20.55 =
abolished. The distance between two axles of metro bogies is
25.11%. On all tracks with other fasteners whose vertical stiffness values
were 20–48 MN/m, the occurrence rate of rail corrugation was equal to
Table 1
1.25/33.14 = 3.77%. Therefore, it could be concluded that the tracks
Rail gauge in China main-line tracks.
with Cologne-egg fasteners more easily suffer from rail corrugation. In
Radius (R) of curved tracks(m) Rail gauge (mm) fact, most of China’s metro tracks have not been equipped with Cologne-
R ≥ 350 1435 egg fasteners. Thus, the authors think that the occurrence rate of rail
350>R ≥ 300 1440 corrugation on most of China’s metro tracks is less than 11.94%.
R < 300 1450
The field investigation into rail corrugation is a very hard task. Based

2
G. Mei and G. Chen Wear 523 (2023) 204727

Fig. 1. Types of railway track beds in China: (a) ballast track bed, (b) monolithic track bed.

Fig. 2. Photographs of rail corrugation occurring on railway main-line tracks: (a) on a curved track of radius 300 m, (b) on a high-speed train track.

Experienced researchers could roughly identify whether rail has suffered

Table 2
from corrugation based on the wheel/rail noise. The authors evaluate
Rail gauge in China metro tracks.
that the occurrence rate of rail corrugation is about 5–8% in China’s
Radius (R) of curved tracks(m) Rail gauge (mm) metro lines without Cologne-egg fasteners in terms of wheel/rail noises.
A-type car B-type car

R > 200 1435 1435

200≥R > 150 1445 1440 2.3. The regularity of rail corrugation
150≥R > 100 1450 1445

It is well known that both curved tracks and tangential tracks are
on the authors’ field tracing investigation into several metro lines from probably subjected to rail corrugation. That gives the illusion that there
the cold-running test of these metro lines to the regular operation for is no regularity in the rail corrugation. After a deep consideration, the
transporting passengers, the authors have found that after the rail authors think that there is a strong regularity in the rail corrugation. In
corrugation occurred for about two months, the passengers could China, the wear-type tread profiles of all wheels are imposed. The
perceive a disturbing but distinguishing wheel/rail noise when the contact status between a wheel and a rail is always one-point contact.
metro train passed through the track section with rail corrugation. The regularity of rail corrugation for the one-point contact wheel-rail
system is narrated as follows.

Fig. 3. Rail corrugation on a curved track of radius 350 m with a superelevation of 120 mm equipped with Cologne-egg fasteners: (a) on the low rail, (b) on the
high rail.

3
G. Mei and G. Chen Wear 523 (2023) 204727

Fig. 4. Rail corrugation on a curved track of radius 300 m with a superelevation of 120 mm: (a) on the low rail, (b) on the high rail.

Fig. 5. Rail corrugation on the same curved track of radius 350 m with superelevation 120 mm for different fasteners: (a) on the low rail with Cologne-egg fasteners,
(b) on the low rail with DTVII-2 fasteners.

Fig. 6. Rail corrugation on the low rail of two ladder tracks of radius 350 m with a superelevation of 120 mm: (a) with DTVII-2 fasteners, (b) with ZX-2 fasteners.

Fig. 7. Rail corrugation on a curved track of radius 350 m with a superelevation of 120 mm: (a) on the low rail, (b) on the high rail.

(1) Whether they are metro lines or railway main-lines, as long as the of the curved tracks of radius 400 m, and developed to the
radii of the curved tracks are less than or equal to 350 m, almost entrance transition curve and the exit transition curve for about
all low rails of the curved tracks will incur corrugation. A curved 10–15 m. If the distance from the stop station to the entrance
track generally consists of an entrance transition curve, a curved transition curve, and the distance from another stop station to the
track with a constant radius, and an exit transition curve. If the exit transition curve are too short, about 10–20 m or less, rail
distance from the stop station to the entrance transition curve, corrugation will cover a partial low rail of the curved tracks with
and the distance from another stop station to the exit transition a constant radius, not develop to the entrance transition curve
curve are enough long, about 200–250 m or more, rail corruga­ and the exit transition curve. In this case, corrugation mainly
tion will cover all length of the low rail of the curved tracks with a covers a center part of the low rail length. For example, rail
constant radius, even develop to some sections of the entrance corrugation shown in Fig. 9 only covered a partial length of the
transition curve and the exit transition curve. For example, rail low rail of the curved tracks of radius 350 m, and not developed
corrugation shown in Fig. 10 has covered all length of the low rail to the entrance transition curve and the exit transition curve.

4
G. Mei and G. Chen Wear 523 (2023) 204727

Fig. 8. Rail corrugation on a curved track of radius 350 m with a superelevation of 120 mm: (a) on the low rail with ZX-2 fasteners, (b) on the low rail with DTVII-
2 fasteners.

Fig. 9. Rail corrugation on a curved track of radius 350 m with a superelevation of 115 mm: (a) on the low rail, (b) on the high rail.

Fig. 10. Rail corrugation on a curved track of radius 400 m with a superelevation of 120 mm: (a) on the low rail, (b) on the high rail.

Anyway, it can be concluded that whether they are metro lines or rail corrugation is an unpredictable event on two rails of the
=railway main-lines, as long as the radii of the curved tracks are curved tracks including tangential tracks.
less than or equal to 350 m, rail corrugation is a deterministic
event on the low rail of the curved tracks. 3. Study on the root cause for rail corrugation
(2) Whether they are metro lines or railway main-lines, as long as the
radii of the curved tracks are larger than 650 m, two rails of the 3.1. Study on the cause to induce rail corrugation on the tightly curved
curved tracks including tangential tracks will seldom incur track
corrugation. Taking into account the fact that the occurrence rate
of rail corrugation is about 1.18–2.73% for railway main-lines, In Section 2, the regularity of rail corrugation has been described:
and 3.77% for metro lines without Cologne-egg fasteners, it can whether they are metro lines or railway main-lines, as long as the radii of
be inferred that the occurrence rate of rail corrugation is less than the curved tracks are less than or equal to 350 m, almost all low rails of
about 1.18–3.77% for the curved tracks whose radii are larger the curved tracks will incur corrugation. Under the condition, rail
than 650 m, including tangential tracks. Therefore, it can be corrugation is a deterministic event on the low rails of the curved tracks.
concluded that if the radii of curved tracks are larger than 650 m, However, whether they are metro lines or a railway main-lines, as long

5
G. Mei and G. Chen Wear 523 (2023) 204727

Fig. 11. Rail corrugation on a curved track of radius 400 m with a superelevation of 120 mm: (a) on the low rail, (b) on the high rail.

Fig. 12. Rail worn scars in a curved track of radius 450 m with a superelevation of 90 mm: (a) on the low rail, (b) on the high rail, (c) on the low rail with light
corrugation.

Fig. 13. Rail corrugation on a tangential track: (a) on the left rail, (b) on the right rail, (c) corrugation initiation in the transition section from booted sleepers to
Cologne-egg fasteners.

6
G. Mei and G. Chen Wear 523 (2023) 204727

Fig. 14. Rail corrugation on a curved track of radius 1100 m with a superelevation of 40 mm: (a) on the low rail, (b) on the high rail.

Fig. 15. Rail scars without rail corrugation on a curved track of radius 600 m with a superelevation of 70 mm: (a) on the low rail, (b) on the high rail.

Fig. 16. Rail scars without rail corrugation on a curved track of radius 800 m with a superelevation of 70 mm: (a) on the low rail, (b) on the high rail.

Fig. 17. Rail scars without rail corrugation on a tangential track: (a) on the left rail, (b) on the right rail.

as the radii of curved tracks are larger than 650 m, two rails of the occurrence rate of rail corrugation is correlated with the saturation state
curved tracks including tangential tracks will seldom incur corrugation. of the creep force between a wheel and a rail. Fig. 19 depicts the vari­
Under the condition, rail corrugation is an unpredictable event on two ation of the saturation coefficient of the creep forces between the leading
rails of the curved tracks including tangential tracks. Fig. 18 shows the wheelset/rails with the radius of curved tracks, which was defined as the
variation of the occurrence rate of rail corrugation with the radius of resultant of the creep force divided by the friction force. The saturation
curved tracks. It is found that when the radius of curved tracks is equal to coefficients of the creep forces were calculated based on SIMPACK
450–650 m, the occurrence rate of rail corrugation dropped sharply simulations of vehicle curve negotiations. It is found that when the
from 100% to 2–4%. The authors think that the sharp drop of the radius of curved tracks is 250–450 m, the saturation coefficients of the

7
G. Mei and G. Chen Wear 523 (2023) 204727

vehicles, tracks and operation parameters. In the succeeding work, the

authors will try to establish a complete boundary to identify whether the
creep forces between wheels and rails are saturated under different
vehicles, tracks, and operation parameters.
When the creep force between a wheel and a rail is saturated, this
saturated creep force can induce a self-excited vibration of the wheel-rail
system. Chen et al. [29] have studied this self-excited vibration and
regarded it as a wavelength-fixing mechanism for rail corrugation.
Subsequently, Chen and his colleagues carried out much research on the
self-excited vibration of the wheel-rail system as a wavelength-fixing
mechanism for rail corrugation [36–39]. Several researchers also
advocated this mechanism [40,41]. In Ref. [29], the friction-induced
self-excited vibration equation of a wheelset-track system was intro­
duced comprehensively. The self-excited vibration of the wheel-rail
system is actually a resonant vibration of the wheel-rail system [42].
The self-excited vibration of the wheel-rail system as a
wavelength-fixing mechanism is an extension of the Grassie and
Kalousek’ corrugation mechanism, that is, the excitation of the track
stiffness by the unsprung mass of the vehicle is the most common
wavelength-fixing mechanism [1]. The authors established a
friction-induced self-excited vibration model of a wheelset-track system.
Fig. 18. Variation of the occurrence rate of rail corrugation.
When the creep forces between the wheels and rails are saturated, there
are always some unstable vibration modes as shown in Fig. 20. One can
creep forces of the leading wheelset/rails for each bogie were larger than see that in the tightly curved track, rail corrugation most likely occurs on
0.95–1.008, in which the creep force between the wheel and rail is the low rail, not on the high rail. This conclusion is consistent with the
thought to be saturated, that is, it is equal to the normal force times the actual field corrugation. It is very interesting that for the wheelset-track
friction coefficient. It is also found that the creep forces of the trailing system, changing the axle-box loadings, rail gauge, rail cant, sleeper
wheelset/rails for each bogie are always unsaturated. Comparing Fig. 18 spacing, stiffness and damping of fasteners, and stiffness and damping of
with Fig. 19, one can see that the radius range corresponding to the sleeper supports, etc. can not eliminate all unstable vibration modes,
sharp drop of the occurrence rate of rail corrugation is approximately in suggesting changing these parameters cannot eliminate rail corrugation.
the vicinity of the radius range corresponding to the boundary from the This conclusion is consistent with the actual field corrugation.
saturation of creep forces to the non-saturation of creep forces. It is
noticed that the boundary from the saturation of creep forces to the non-
saturation of creep forces was located at 400–450 m as shown in Fig. 19, 3.2. Study on the cause to induce rail corrugation on the mildly curved or
while the boundary corresponding to the sharp drop of the occurrence tangential tracks
rate of rail corrugation was evaluated to be located at 450–650 m. There
is a difference between two boundaries. The authors took account into a According to the authors’ field investigation result and other re­
safety margin. As we know, a SIMPACK simulation is a numerical searchers’ reports, rail corrugation sometimes occurs on smoothly
simulation, and is also an approximate simulation. The authors took curved tracks whose curve radii are larger than 450–650 m or on
radius 650 m as a change point at which the occurrence rate of rail tangential tracks. In Section 2, the authors roughly estimated that the
corrugation dropped abruptly. It is because when the radius of curved occurrence rate of rail corrugation on China’s railway main-lines or
tracks is equal to or larger than 650 m, the creep forces between the metro lines without Cologne-egg fasteners is less than 3.77%. Re­
leading wheelset/rails are almost always unsaturated. It needs searchers need to understand why the occurrence rate of rail corrugation
mentioning that nowadays the authors use SIMPACK package to simu­ on China’s railway main-lines or metro lines without Cologne-egg fas­
late the vehicle-track system dynamics to identify whether the creep teners is so low compared with the occurrence rate of rail corrugation on
forces between wheels and rails are saturated or not. However, SIM­ tightly curved tracks, and what causes this corrugation. In the smoothly
PACK simulations of vehicle-track systems still need to spend much curved tracks or tangential tracks, the creep forces between wheels and
time. Fig. 19 depicts a boundary for identifying whether the creep forces rails are generally unsaturated. Under the extreme condition, however,
between wheels and rails are saturated or not. It is only valid for specific the creep forces between wheels and rails are probably saturated. For

Fig. 19. Variation of the saturation coefficient of the creep force vs the radius of curved tracks: (a) for type-A vehicle, (b) for type-B vehicle.

8
G. Mei and G. Chen Wear 523 (2023) 204727

Fig. 20. Unstable vibrations of the wheelset-track system with a radius of 350 m and a superelevation of 115 mm: (a) distribution of unstable vibrations, (b) an
unstable vibration mode of frequency 223.04 Hz and effective damping ratio − 0.0175.

example, the creep forces between wheels and rails may be saturated in that spray sand onto the rails when necessary to suppress the slippage of
the serious traction and braking applications of trains. It is well known the driving wheels on the rails.
that in the design of the train traction and braking, the slip of wheels on Similarly, the aim of checking the wheel/rail slip in braking appli­
rails must be checked and is not allowed to occur. Fig. 21 depicts an cations is also to avoid the slip of wheels on rails. Fig. 22 depicts the
adhesion curve of traction forces. When the train speed is less than about variation of the braking wheel/rail adhesion factor against speed. It is
53 km/h, the traction acceleration of trains is limited by the adhesion observed that the braking adhesion factor changes in a large range. After
curve. That is to say, in the speed range of 0–53 km/h, the traction ac­ the braking slip is checked, the braking slip can be avoided in most in­
celeration of trains cannot be too large, otherwise the driving wheels stances. However, the braking slip probably takes place under the
may slip on the rails. Many parameters affect the adhesion curve of extreme condition. In railway main-lines, the distance between two stop
traction forces. Air temperature, humidity, rain, dust, snow, etc. all stations is 15–100 km. The braking deceleration of trains is about
affect the adhesion curve. After the traction slip is checked, the traction 0.5–0.8 m/s2. In metro lines, most of the distances between two stop
slip can be avoided in most instances. The traction acceleration is about stations are 800–1200 m. The braking deceleration of metro trains is
0.4–0.8 m/s2 for railway main-line trains, 0.6–1.0 m/s2 for metro trains. about 0.8–1.4 m/s2. Therefore, subway train wheels are more likely to
The greater the traction acceleration, the more likely the wheels slip on slip on the rails than main-line train wheels. This explains why the
the rails. The traction slip probably takes place during train operation. mildly curved or tangential tracks of metro lines are more prone to
For example, all locomotives are equipped with sandblasting machines corrugation than those of railway main-lines. There is an example to
support the viewpoint that braking applications can suppress or remove
rail corrugation. China’s high-speed trains and metro trains are all
equipped with anti-slip devices to minimize the slip of wheels on the
rails in traction and braking applications. The anti-slip device controls
the slip of wheels on the rails based on the difference between the
rotational speeds of wheelset axles or the maximum traction accelera­
tion and braking deceleration. Obviously, the control thresholds of the
rotational speed difference and the maximum traction acceleration or
braking deceleration are not unique values, which affect the slip of
wheels on the rails in tribological sense.
Once the wheels slip on the rails, the slip friction can induce self-
excited vibrations of the wheelset-track system, subsequently, rail
corrugation. Fig. 23 shows a distribution of unstable vibrations of the

Fig. 21. Adhesion curve of traction forces. Fig. 22. Variation of the measured adhesion factor with speed.

9
G. Mei and G. Chen Wear 523 (2023) 204727

Fig. 23. A distribution of unstable vibrations of the wheelset-track with a radius of 1100 m and a superelevation of 40 mm: (a) distribution of unstable vibration, (b)
an unstable vibration mode of frequency 250.72 Hz and effective damping ratio − 0.0102.

wheelset-track system and an unstable vibration mode. One can see that was 21,000–25,000 kg. About 110,000 km of railway main-lines were
the slip friction can induce self-excited vibrations of the wheelset-track equipped with ballast-track beds. The irregularities of the railway main-
system and rail corrugation. Fig. 24 shows predicted possible rail line tracks with ballast-track beds are greater than those of the metro
corrugation on a tangential track. One can observe that once the wheels tracks with monolithic track beds. In China, the main specification pa­
slip on the rails, the slip friction can induce rail corrugation, and rail rameters of the subway tracks are similar to those of the railway main-
corrugations will suffer from two rails. line tracks, including rail type, rail pad, rail gauge, inclination, super­
It needs to be mentioned that in Fig. 23, the resultants of the creep elevation and so on.
forces between the wheels and rails are close to the longitudinal direc­ Obviously, the wheel-rail interaction of the railway main-lines is
tion of the track, while in Fig. 20, the resultants of the creep forces be­ stronger than that of the subway lines. However, the occurrence rate of
tween the wheels and rails are close to the lateral direction of the track, rail corrugation on the railway main-lines is smaller than that on the
which depend on the SIMPACK simulation results. In Fig. 24, the re­ subway lines. According to the authors’ field investigations in China, a
sultants of the creep forces between the wheels and rails are almost 30–50 km long metro line always incurs rail corrugation to some extent,
along the longitudinal direction of the track. while a 50–100 km long railway main-line rarely incurs rail corrugation.
The authors attribute this phenomenon to wheel slippage on rails.
Whether it is a subway line or a railway main-line, as long as the radius
3.3. Discussion of the curved track is less than or equal to 350 m, the wheels of the
leading wheelset of each bogie always slip on the rails. The friction
3.3.1. On the generation mechanism of rail corrugation on mildly curved caused by the wheel-rail slippage can cause self-excited vibration of the
tracks or tangential tracks wheelset-track system, resulting in rail corrugation. Since the contact
Interestingly, the occurrence rate of rail corrugation on main railway angle between the wheel of the leading wheelset and high rail reaches
lines is very close to that of subway lines without the Cologne-egg fas­ 20–35◦ and the contact angle between the wheel of the leading wheelset
teners. The former is about 1.18–2.73%, and the latter is about 3.77%. In and low rail is about 2.5◦ , the saturated creep force can only induce self-
China, the vehicle and track quality of subway lines is superior to that of excited vibration of the wheel and low rail system depending on the
main railway lines. All metro vehicles were equipped with primary wheel-rail contact angle under the saturation of creep forces [37].
suspension springs and dampers, as well as secondary suspension springs Therefore, the low rails of the sharply curved tracks most likely incur
and dampers. All subway tracks were equipped with monolithic track corrugation. Whether it is a subway line or a railway main-line, as long
beds. The maximum allowable axle load for these metro vehicles was as the radius of the curved track is less than or equal to 350 m, the
14,000–16,000 kg. By the end of 2021, the railway main-lines held occurrence rate of rail corrugation is close to 100%. This is attributed to
966,000 freight cars and 78,000 passenger cars. All freight cars were the inherent saturation of the creep forces of the leading wheelset in
equipped with three-piece bogies without primary suspension springs sharply curved tracks with a radius of 350 m or less. Whether the vehicle
and dampers. The maximum allowable axle load for these freight cars

Fig. 24. A distribution of the unstable vibrations of the wheelset-track on a tangential track: (a) distribution of unstable vibration, (b) an unstable vibration mode of
frequency 246.91 Hz and effective damping ratio − 0.0143.

10
G. Mei and G. Chen Wear 523 (2023) 204727

negotiates a sharply curved track under coasting, driving or braking changed from 85 km/h to 65 km/h in 8 s. Due to excessive braking
conditions, the creep forces of the leading wheelset are always saturated. deceleration of subway trains, the wheel slippage was most likely to
However, whether it is a subway line or a railway main-line, as long as occur. Liu also reported that when the braking application was removed
the radius of the mildly curved track is greater than or equal to 650 m, by adjusting the metro train operation speed, rail corrugation did not
including the tangential track, the occurrence rate of rail corrugation reappear on the ground rail.
will drop sharply to 3–4%. This is attributed to the inherent unsaturation
of the creep forces of the leading and trailing wheelsets in the mildly 3.3.2. Suppression of rail corrugation on mildly curved tracks or tangential
curved track with a radius greater than or equal to 650 m or in the tracks
tangential track. When vehicles go through the mildly curved tracks In China, almost every subway line suffers from more or less rail
under the coasting condition, the creep forces of the wheelsets are al­ corrugation. Suppressing and eliminating rail corrugation is an urgent
ways unsaturated. When vehicles go through the mildly curved tracks task. According to the authors’ field investigation, whether they are
under most of the driving or braking conditions, the creep forces of the metro lines or railway main-lines, as long as the radii of the curved
wheelsets are generally unsaturated. This is because wheel-rail slippage tracks are less than or equal to 350 m, the occurrence rate of rail
must be checked to avoid wheel-rail slippage before determining the corrugation is very close to 100%. As long as the radii of the curved
traction acceleration or braking deceleration of the vehicles. These tracks are greater than or equal to 650 m, the occurrence rate of rail
checks can indeed avoid most of the wheel-rail slippages. Under the corrugation sharply drops to 3–4%. Eliminating rail corrugations on
extreme condition, however, the wheel-rail slippage is possible because tightly curved rails is a difficult task. In the literature, three passive
of the complex wheel-rail contact condition. Of course, this slippage is suppression methods have been proposed, including grinding rails,
very limited. Compared with railway main-lines, metro lines are more using friction modifiers, and changing travelling speeds [2]. Mei et al.
prone to wheel-rail slippage. This is attributable to the fact that the [44] have proposed an active suppression method to remove rail
traction acceleration or braking deceleration of subway trains is greater corrugation on the tightly curved tracks. Until there are better ways to
than that of main-line trains. At present, China’s subway trains are actively control rail corrugation on the tightly curved tracks, researchers
generally controlled by the automatic train operation (ATO) mode. In may consider Mei’s approach to actively eliminate rail corrugation on
the ATO mode, when the trains are driven or when the trains are braked the tightly curved tracks.
is all accurate. That is to say, once a train has wheel-rail slippage in a In the present paper, the authors have proposed the slip of wheels on
certain track section, every train will always have wheel-rail slippage in rails as the root cause of rail corrugation. This new insight may be used
the same track section. Naturally, rail corrugation is more easily created to control rail corrugation on mildly curved tracks or tangential tracks.
in the subway track section with wheel-rail slippage. Fig. 25 depicts the In order to suppress or remove rail corrugation on mildly curved tracks
variation of the measured electric current passing through one panto­ or tangential tracks, two measures need to be taken. Firstly, the corru­
graph with the distance between two metro stop stations. The electric gated rail must be ground cleanly. Secondly, the traction acceleration or
braking of the metro train was easily identified based on a negative braking deceleration needs to be decreased, even be cancelled in the
electric current. The electric braking can create a great brake force at the track section with rail corrugation. The author believes that by taking
wheel-rail interface. It needs to be mentioned that electric braking is these two measures, rail corrugation on slightly curved tracks or
generally applied at a high train speed. Coincidentally, as reported by tangential tracks will be greatly suppressed or even eliminated.
Zhang et al. [35], rail corrugations generally appear in the track sections
of mildly curved tracks where train speeds are higher. At present, the 4. A fast prediction method for rail corrugation
authors lack the field survey data on the correlation between the rail
corrugation on the gently curved tracks and the electric braking of The ultimate goal of studying rail corrugation is to be able to accu­
subway trains. The correlation will be studied in the future work. rately predict where rail corrugation will occur in all new railway lines
There is one example to support the authors’ viewpoint that braking at the design stage, and where rail corrugation has occurred and where
applications probably cause rail corrugation. Liu [43] reported that the rail corrugation has not occurred in old railway lines assuming that the
low rail of a curved track with a radius of 700 m and a superelevation of information on rail corrugation occurrence is kept confidential. In the
119 mm suffered from corrugation. Liu found that in the track section literature, many researchers reported that they had successfully verified
with rail corrugation, the subway train was braked, and the train speed their models. Chen, the second author of the present paper, pointed out
that there is a flaw in their model validations [36], and proposed a
benchmark condition for model validations. Here, the occurrence rate of
track corrugation on metro lines and railway main-lines can be used as a
validation condition for the verification of various rail corrugation
prediction models. As mentioned in Sections 2.1 and 2.2, only 280 km of
the 10,257 km railway main-line tracks incurred rail corrugation; only
6.4 km of 53.69 km subway tracks incurred rail corrugation. The re­
searchers encountered a challenge in identifying where 280 km corru­
gated tracks occurred in 10,257 km main-line tracks, and where 6.4 km
corrugated tracks occurred in 53.69 km subway tracks, using prediction
models published in the literature. Many models published in the liter­
ature can always successfully predict rail corrugation, but are difficult to
identify the conditions without rail corrugation.
According to our research experience, the authors propose the slip of
wheels on rails as the root cause for rail corrugation. Without the slip, no
corrugation occurs. Whether it is a main-line track or a metro track,
whether it is a sharply curved track, a mildly curved track, or a
tangential track, rail corrugation will be generated in the track section
where the wheel slips on the rail. The slip friction can induce self-excited
vibrations of the wheelset-track system.
Fig. 25. Variation of the measured train speed and electric current with the Establishing a prediction model shown in Fig. 20b will spend 3–6
travelling distance. months. It needs much time consumption. In the paper, the authors

11
G. Mei and G. Chen Wear 523 (2023) 204727

propose a fast prediction method for rail corrugation. It is narrated as 1. The field investigation into rail corrugation shows that the occur­
follows: Firstly, using SIMPACK to identify whether the creep force be­ rence rates of rail corrugation on the China’s metro lines and railway
tween the wheel and rail is saturated based on the radius of curved main-lines are 11.94% and 2.73%, respectively. These two occur­
tracks. If the creep force is saturated, rail corrugation is considered to rence rates of rail corrugation can be used as the benchmark condi­
occur. If the creep force between the wheel and rail is not saturated, rail tions for rail corrugation model validations.
corrugation is considered not to occur. If the radius of curved tracks is 2. The slip of wheels on rails is proposed as the root cause for rail
larger than 650 m, including tangential tracks and without brake and corrugation occurrence. Without the slip, no corrugation occurs.
traction applications, the creep force must be unsaturated. In this case, 3. Rail corrugation on the mildly curved tracks or tangential tracks can
we can judge that no rail corrugation occurs. The accuracy of the fast be suppressed or eliminated by changing braking or traction condi­
prediction method for rail corrugation is larger than 85–90% or more. tions to remove the slip of wheels on rails.
Encouragingly, using the fast prediction method to evaluate the occur­ 4. The fast prediction method for rail corrugation may be used to pre­
rence rates of rail corrugation on the China’s metro lines and railway dict rail corrugation occurrence with a high accuracy.
main-lines, we can obtain a high accuracy of 85–90% or more. For
example, we know that the radii of curved tracks in China’s railway Declaration of competing interest
main-lines are greater than 800–1000 m. We also know that under the
curved tracks with radii of more than 650 m, the wheels will not slip on The authors declare that they have no known competing financial
the rails. Therefore, we can roughly estimate that the occurrence rate of interests or personal relationships that could have appeared to influence
rail corrugation is zero. The error between the fast predicted occurrence the work reported in this paper.
rate and the actual occurrence rate is 2.73%-0 = 2.73%. It’s very small.
The prediction accuracy reaches 97.27%. For another example, we use Acknowledgements
the fast prediction method to identify the length of the corrugated tracks
of 53.69 km subway tracks. We know that the accumulated length of the The authors thank the financial supports from National Natural
sharply curved tracks whose radii are less than 400 m is 4.18 km. Ac­ Science Foundation of China (No. 52175189).
cording to Fig. 19, when metro vehicles negotiate curved tracks with
radius 400 m or less, the wheels will slip on the rails. Thus, we can References
roughly evaluate that the occurrence rate of rail corrugation is equal to
4.18/53.69 = 7.79%. The error between the fast prediction occurrence [1] S.L. Grassie, J. Kalousek, Rail corrugation: characteristics, causes and treatments,
Proc. Inst. Mech. Eng. - Part F J. Rail Rapid Transit 207 (1) (1993) 57–68.
rate and the actual occurrence rate is equal to 11.94%–7.79% = 4.15%. [2] S.L. Grassie, Rail Corrugation, Wheel–Rail Interface Handbook, CRC Press, 2009,
The prediction accuracy reaches 95.85%. pp. 349–376.
Although the occurrence rate of rail corrugation on metro mild curve [3] J. Kalousek, K. Johnson, An investigation of short pitch wheel and rail corrugations
on the Vancouver mass transit system, Proc. Inst. Mech. Eng. - Part F J. Rail Rapid
tacks and straight tracks is very small, less than 3–4%, rail corrugation is Transit 206 (2) (1992) 127–135.
inevitable. This is attributed to large braking and traction forces which [4] K. Hempelmann, K. Knothe, An extended linear model for the prediction of short
lead to the saturation of the creep forces between the wheels and rails. pitch corrugation, Wear 191 (1) (1996) 161–169.
[5] J.C.O. Nielsen, N.R. Lund, A. Johansson, Train-track interaction and mechanisms of
The authors think that if the traction acceleration or braking decelera­ irregular wear on wheel and rail surfaces, Veh. Syst. Dyn. 40 (1–3) (2003) 43–54.
tion is less than 0.6 m/s2, the track section will not incur rail corruga­ [6] J. Gomez, E.G. Vadillo, J. Santamaria, A comprehensive track model for the
tion, otherwise the track section will probably incur rail corrugation. improvement of corrugation models, J. Sound Vib. 293 (3) (2006) 522–534.
[7] P.A. Meehan, W.J.T. Daniel, T. Campey, Prediction of the growth of wear-type rail
Especially, when the traction acceleration or braking deceleration is
corrugation, Wear 258 (7–8) (2005) 1001–1013.
greater than 1.0 m/s2, the track section will most likely incur rail [8] T.X. Wu, D.J. Thompson, An investigation into rail corrugation due to micro-slip
corrugation. Of course, this conclusion needs further verification. under multiple wheel/rail interactions, Wear 258 (7–8) (2005) 1115–1125.
In the design stage of new railway lines, the parameters of the ve­ [9] Y. Sato, A. Matsumoto, K. Knothe, Review on rail corrugation studies, Wear 253
(1–2) (2002) 130–139.
hicles, tracks and operating conditions have been determined. There­ [10] K.H. Oostermeijer, Review on short pitch rail corrugation studies, Wear 265 (9–10)
fore, vehicle dynamics parameters, such as wheel-rail contact angle, (2008) 1231–1237.
wheel-rail contact normal force, suspension force acting on the axle [11] J. Kalousek, K.L. Johnson, An investigation of short pitch wheel and rail
corrugations on the Vancouver mass transit system, J. Rail Rapid Transit 206
box, wheel-rail creep force, etc., can be obtained by SIMPACK simula­ (1992) 127–135.
tions of the vehicle-track systems. Based on these vehicle dynamics [12] P.A. Meehan, P.A. Bellette, R.D. Batten, W.J.T. Daniel, R.J. Horwood, A case study
parameters, the designers can check whether the wheel-rail creep force of wear-type rail corrugation prediction and control using speed variation,
J. Sound Vib. 325 (2009) 85–105.
is saturated or not, then judge where rail corrugation will occur in new [13] A.R. Valdivia, A linear dynamic wear model to explain the initiating mechanism of
railway lines using the present fast prediction method of rail corruga­ corrugation, Veh. Syst. Dyn. 17 (sup1) (1988) 493–496.
tion. As above mentioned, the prediction accuracy of the fast prediction [14] E. Tassilly, N. Vincent, A linear model for the corrugation of rails, J. Sound Vib.
150 (1) (1991) 25–45.
method for rail corrugation reaches 85–90%, or even higher. Thus, the [15] S. Muller, A linear wheel-track model to predict instability and short pitch
authors believe that the root cause of the rail corrugation, the wheel corrugation, J. Sound Vib. 227 (5) (1999) 899–913.
slipping on the rail, may have been found. Once the locations of rail [16] A. Igeland, H. Ilias, Rail head corrugation growth predictions based on non-linear
high frequency vehicle/track interaction, Wear 213 (1) (1997) 90–97.
corrugation occurrence are identified, active and passive precautions
[17] P.T. Torstensson, J.C.O. Nielsen, Simulation of dynamic vehicle-track interaction
can be applied to prevent the rail corrugation. The active precautions on small radius curves, Veh. Syst. Dyn. 49 (11) (2011) 1711–1732.
include an application of the wheel construction without rail corruga­ [18] X. Sheng, D.J. Thompson, C.J.C. Jones, G. Xie, S.D. Iwnicki, P. Allen, S.S. Hsu,
tion on sharp curves [44], and decreasing the acceleration values of Simulations of roughness initiation and growth on railway rails, J. Sound Vib. 293
(3–5) (2006) 819–829.
braking and traction applications to eliminate the wheel slip on the rails [19] L. Baeza, P. Vila, G. Xie, S.D. Iwnicki, Prediction of rail corrugation using a rotating
in mild curves and straight tracks. The passive precautions include the flexible wheelset coupled with a flexible track model and a non-Hertzian/non-
application of rail grinding, friction modifiers and variable passing steady contact model, J. Sound Vib. 330 (18–19) (2011) 4493–4507.
[20] A. Luciano, C. Michele, Short pitch corrugation of railway tracks with wooden or
speeds [12,45]. concrete sleepers: an enigma solved? Tribol. Int. 43 (2010) 610–622.
[21] P.A. Bellette, P.A. Meehan, W.J.T. Daniel, Contact induced wear filtering and its
5. Conclusions influence on corrugation growth, Wear 268 (2010) 1320–1328.
[22] Y.R. Wang, T.X. Wu, The growth and mitigation of rail corrugation due to
vibrational interference between moving wheels and resilient track, Veh. Syst. Dyn.
In the paper, field investigations into rail corrugation on the China’s 58 (8) (2020) 1257–1284.
metro lines and railway main-lines with several theoretical simulations
were performed. The following several conclusions can be drawn.

12
G. Mei and G. Chen Wear 523 (2023) 204727

[23] C.Z. Ma, L. Gao, T. Xin, X.P. Cai, M.M. Nadakatti, P. Wang, The dynamic resonance [35] H.G. Zhang, W.N. Liu, W.F. Liu, Z.Z. Wu, Study on the cause and treatment of rail
under multiple flexible wheelset-rail interactions and its influence on rail corrugation for Beijing metro, Wear 317 (2014) 120–128.
corrugation for high-speed railway, J. Sound Vib. 498 (2021), 115968. [36] G.X. Chen, S. Zhang, B.W. Wu, X.N. Zhao, Z.F. Wen, H. Ouyang, M.H. Zhu, Field
[24] X.S. Jin, Z.F. Wen, K.Y. Wang, Z.R. Zhou, Q.Y. Liu, C.H. Li, Three-dimensional measurement and model prediction of rail corrugation, Proc IMechE Part F: J. Rail
train–track model for study of rail corrugation, J. Sound Vib. 293 (2006) 830–855. Rapid Transit 234 (4) (2020) 381–392.
[25] X.G. Liu, P. Wang, Investigation of the generation mechanism of rail corrugation [37] X.L. Cui, G.X. Chen, H.G. Yang, Q. Zhang, H. Ouyang, M.H. Zhu, Effect of the
based on friction induced torsional vibration, Wear (2021) 468–469, 203593. wheel/rail contact angle and the direction of the saturated creep force on rail
[26] C.A. Brockley, P.L. Ko, An investigation of rail corrugation using friction-induced corrugation, Wear 330 (2015) 554–562.
vibration theory, Wear 128 (1) (1988) 99–105. [38] X.L. Cui, G.X. Chen, H. Ouyang, Study on the effect of track curve radius on
[27] Y. Suda, M. Hanawa, M. Okumura, T. Iwasa, Study on rail corrugation in sharp friction-induced oscillation of a wheelset–track system, Tribol. Trans. 62 (4) (2019)
curves of commuter line, Wear 253 (1) (2002) 193–198. 688–700.
[28] Y.Q. Sun, S. Simson, Wagon-track modelling and parametric study on rail [39] X.L. Cui, G.X. Chen, H.G. Yang, Q. Zhang, H. Ouyang, M.H. Zhu, Study on rail
corrugation initiation due to wheel stick-slip process on curved track, Wear 265 corrugation of a metro tangential track with Cologne-egg type fasteners, Veh. Syst.
(9–10) (2008) 1193–1201. Dyn. 54 (3) (2016) 353–369.
[29] G.X. Chen, Z.R. Zhou, H. Ouyang, X.S. Jin, M.H. Zhu, Q.Y. Liu, A finite element [40] O. El Beshbichi, C. Wan, S. Bruni, E. Kassa, Complex eigenvalue analysis and
study on rail corrugation based on saturated creep force-induced self-excited parameters analysis to investigate the formation of railhead corrugation in sharp
vibration of a wheelset–track System, J. Sound Vib. 329 (2010) 4643–4655. curves, Wear (2020) 450–451, 203150.
[30] P.T. Torstensson, M. Schilke, Rail corrugation growth on small radius curves- [41] D. Fourie, R. Frohling, S. Heyns, Railhead corrugation resulting from mode-
Measurements and validation of a numerical prediction model, Wear 303 (1–2) coupling instability in the presence of veering modes, Tribol. Int. 152 (2020),
(2013) 381–396. 106499.
[31] M. Hiensch, J.C.O. Nielsen, E. Verheijen, Rail corrugation in The Netherlands- [42] N.M. Kinkaid, O.M. O’ Reilly, P. Papadopoulos, Automotive disc brake squeal,
measurements and simulations, Wear 253 (1–2) (2002) 140–149. J. Sound Vib. 267 (2003) 105–166.
[32] E. Tassilly, N. Vincent, Rail corrugations: analytical model and field tests, Wear [43] S.Q. Liu, Analysis and solution of abnormal rail corrugation in a metro line,
144 (1–2) (1991) 163–178. Railway Signal Commun 50 (5) (2014) 18–19 (in Chinese).
[33] R.D. Batten, P.A. Bellette, P.A. Meehan, R.J. Horwood, W.J.T. Daniel, Field and [44] Mei, G.M.; Chen, G.X.; Yan, S.; Chen, R.X., Study on a heuristic wheelset structure
theoretical investigation of the mechanism of corrugation wavelength fixation without rail corrugation on sharply curved tracks. Shock Vib. (2021) 3874005.
under speed variation, Wear 271 (1–2) (2011) 278–286. [45] D. Eadie, J. Kalousek, K.C. Chiddick, The role of high positive friction modifier in
[34] N. Correa, E.G. Vadillo, J. Santamaria, J. Herreros, A versatile method in the space the control of short pitch corrugations and related phenomena, Wear 253 (2002)
domain to study short-wave rail undulatory wear caused by rail surface defects, 185–192.
Wear 352 (1) (2016) 196–208.

13