See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/337892966

Evaluation of derailment risk by simplified analytic procedure and computer

simulation

Conference Paper in AIP Conference Proceedings · December 2019

DOI: 10.1063/1.5138333

CITATION READS
1 531

3 authors:

Yunendar Handoko Yunus Sinaga

Bandung Institute of Technology Bandung Institute of Technology
9 PUBLICATIONS 83 CITATIONS 1 PUBLICATION 1 CITATION

SEE PROFILE SEE PROFILE

Prasetya Adi Nugraha

PT INKA (Persero)
1 PUBLICATION 1 CITATION

SEE PROFILE

All content following this page was uploaded by Prasetya Adi Nugraha on 29 December 2020.

The user has requested enhancement of the downloaded file.

 Evaluation of Derailment Risk by Simplified Analytic
Procedure and Computer Simulation
Yunendar Aryo Handoko1, a), Yunus Sinaga1, b)
and Prasetya Adi Nugraha 2, c)
1
Institut Teknologi Bandung
2
PT Industri Kereta Api (Persero)
a)
Corresponding author: yunendar@ftmd.itb.ac.id, yunendar.handoko@gmail.com
b)
yunussinaga206@gmail.com
c)
prasetya.anugraha@inka.co.id

Abstract. One of the tasks must be done by engineer in the design process of a new rail vehicle is the evaluation of
derailment risk in order to ensure the safety of the vehicle against the derailment accident. Performing such task by time-
series simulation using computer software is time consumed, high cost as well as difficult especially in the early design
stage where only limited information available. It is practical in the rail car manufacture industry that when firstly define
the vehicle configuration a procedure using simplified analytic equations is used to evaluate the sensitivity of the vehicle
against derailment prior to the computer simulation in the later stage. This paper examines the procedure of the
calculation of derailment risk parameters using simplified analytic equations and compare it with the result of computer
simulation of full multibody dynamics model of a railway vehicle.

INTRODUCTION
When passing the curve, the movement of railway vehicle is guided by contact mechanism between the wheel
and the rail so the vehicle can follow the geometric layout of the track [1]. As the vehicle tends to follow its initial
running direction the wheelset will move outward to the outer rail when it comes into the curve track which cause
different rolling radius between the inner and outer wheel. Due to conicity of the wheel, the outer wheel will have
larger rolling radius and the wheelset will naturally turns following the track curve as both wheels are connected by
a rigid axle. In the sharp curve the wheelset moving further outward until the flange of the outer wheel touches the
outer rail and gives bigger turning force as illustrated in Figure 1. However, this situation can lead to the derailment
accident when the lateral force is big enough to force the flange moving further outward and climbs the outer rail.

FIGURE 1. (a) wheelset central position. (b) wheelset position when passing the curve
 In the design stage of a new rail vehicle, it is important to evaluate the derailment risk to ensure the safety of the
vehicle against the derailment accident. Flange climbing derailment is the typical derailment occurred when the rail
vehicle passing the curve. The sensitivity of a railway vehicle against flange climbing derailment can be measured
by Nadal’s criterion [2-5]. If the derailment coefficient is lower than critical value calculated using Nadal’s
equation, then the derailment is justified to be “unlikely happen”. The derailment coefficient is stated in form of the
lateral force (Y) generated at the contact point between the wheel flange and the rail divided by the vertical force
(Q) at the same point as shown in Figure 2.

FIGURE 2. Forces at flange contact point.

According to the Nadal’s formulation the critical value of Y/Q is defined by the flange contact angle and the
wheel-rail friction coefficient as shown in the Equation 1.

𝑌 tan 𝛼−𝜇
( ) = (1)
𝑄 𝑐𝑟𝑖𝑡 1+𝜇 tan 𝛼

For standard wheel profile in Indonesia, which has 67° flange contact angle, and friction coefficient of 0.3 the
critical value of Y/Q is 1.2. However, for the design purpose maximum value of 1.0 is commonly chosen.

CALCULATION OF LATERAL AND VERTICAL FORCE USING

SIMPLIFIED ANALYTIC EQUATION
When the vehicle running through curved track it generates lateral forces in the contact point between the wheel
and the rail. The magnitude of this lateral force depends on several parameters such as curve radius, vehicle speed,
track cant gradient and friction coefficient in the point of contact. At the same time the vertical force on the wheel
may reduce due to twist caused by the track cant gradient and irregularity. This could lead to the increase of Y/Q
value which, according to Nadal’s criterion, represents the increase of the derailment risk.
The total lateral force exerted by the wheel on the outer rail can be calculated using equation 2 [6]:

𝑌𝑡𝑜𝑡 = (𝑌𝑓𝑙𝑎𝑛𝑔𝑒 + 𝑌𝑐𝑒𝑛𝑡𝑟 + 𝑌𝑤𝑖𝑛𝑑 ) + 𝑌𝑑𝑦𝑛 (2)

The first 3 terms in the bracket (flange force, uncompensated centrifugal force and wind force) are considered as
quasi-static forces while the last term is dynamic force due to track irregularity. For a matter of simplification, in this
study we consider only the quasi-static forces except the wind force.
 FIGURE 3. Curving forces.

Figure 3 shows forces acting on wheelsets of a bogie when negotiating curve [7]. From the figure we can see that
the leading wheelset acts as the main guide to direct the movement of the vehicle following the track curve and more
vulnerable to experience wheel-climb derailment. The force on the wheel flange (𝑌𝑓𝑙𝑎𝑛𝑔𝑒 ) of the outer wheel of the
leading wheelset is balanced by the friction force on tread of the inner and outer wheel and produce torque that
change the travel direction of the wheel following the curved track. However, when the outer wheel starts to climb
the rail, the tread of the outer wheel will lose contact with the rail. In such condition, to maintain the quasi-static or
steady state curving, the flange force is limited only by the friction coefficient between the inner wheel tread and the
rail and can be estimated according to Equation 3.

(𝑌𝑓𝑙𝑎𝑛𝑔𝑒 ) = 𝜅𝑄𝑖 (3)

𝑚𝑎𝑥

where 𝑄𝑖 is the inner wheel load and 𝜅 is a parameter depend on the curve radius and wheel-rail friction coefficient
which value was defined by Takai et al [8].
The lateral force due to non-compensated centrifugal force is calculated by the following equation.

𝑣2 𝐶
𝑌𝑐𝑒𝑛𝑡𝑟 = 𝑊0 (𝑔𝑅 − 𝐺 ) (4)

The vertical load on the outer wheel is:

∗
𝑊𝑜 𝑣2 𝐶 𝐻𝐺 𝑣2 𝐶
𝑄𝑜 = {(𝛾 + )+ 𝐺 ( − )} (5)
2 𝑔𝑅 𝐺 𝑔𝑅 𝐺
2

The vertical load on the inner wheel is:

𝑊𝑜 𝑣2 𝐶𝐺 𝐻∗ 𝑣2 𝐶
𝑄𝑖 = {((2 − 𝛾) + 𝑔𝑅 𝐺 ) − 𝐺/2 (𝑔𝑅 − 𝐺 )} (6)
2

where C is cant in curved track, G is gauge, 𝛾 is static wheel load ratio of outside wheel, and 𝐻𝐺∗ is effective height
of center-of-gravity of vehicle (assumed to be 1.25 times the real height of center-of-gravity for non-tilt vehicles
without anti-roll bars). This study targets at vehicles with bolster bogies and quasi-static condition where the lateral
force due to wind and dynamic force due to track irregularity are not put into consideration.
 COMPUTER SIMULATION
The full vehicle model was developed by using multibody dynamics software Universal Mechanism which has a
capability to perform time-series simulation to investigate the vehicle dynamics performances [9]. The model is
shown in Figure 4. The multibody model of the vehicle and the track consists of 11 bodies - including the left and
the right rails which is modelled as separate bodies with 3 d.o.f each (vertical, lateral and rotation about longitudinal
axis). To make the model as close as possible the real vehicle, non-linearities in the suspension elements
characteristics are put into account including the friction on the pivoting disk and bump-stop elements to represent
slack and clearance in the connection. The wheel-rail contact forces are calculated using Kik and Piotrowski method
[10], which improves Kalker’s contact formulation [11], and consider more than one point of contact.

FIGURE 4. Vehicle model.

Figure 5.a shows the typical lateral forces generated between the wheel and the rail during the vehicle passing a
curved track resulted from the computer simulation. As shown in the figure, the vehicle started running in the
straight track and entered the transition curve before negotiating the main curve with constant radius. After passing
the main curve the vehicle entered the second transition curve and then came out from the curve back to the straight
track. In the main curve with constant radius a balance condition was attained and a relatively constant lateral forces
in outer and inner wheel was generated. These lateral forces kept the vehicle to follow the curved track. A similar
pattern applied for the vertical forces as shown in Fig 5. (b).

(a) (b)
FIGURE 5. Lateral force (a) and vertical force (b) between the wheel and the rail during the vehicle passing a curved track
(v = 60km/h)

Knowing lateral force and vertical force generated between the wheel and the rail, we can then calculate the
derailment coefficient as shown in in Figure 6. From the figure the derailment coefficient is around 0.25 during the
steady curve and reach the maximum value of 0.3 when the vehicle come out from the main curving into the
transition zone. The sudden increase of the derailment coefficient in the transition zone is due to the change of the
track cant which reduce the vertical force in the outer wheel.

FIGURE 6. Derailment coefficient Y/Q.

COMPARISON OF SIMPLIFIED ANALYTIC RESULTS

WITH COMPUTER SIMULATION
Using the same vehicle parameters input as in the computer model, the lateral and the vertical forces at the
wheel-rail contact point as well as the derailment coefficient were calculated using simplified analytical formulation
given in section 1. Then the results are compared to those calculated by computer simulation as shown in Figure 7,
which shows the lateral and vertical forces at various speed during negotiating 400 m radius curve. Noted that the
forces calculated using the analytic formulation are the balance forces during the constant radius curve, so they are
compared to the magnitude of the forces in the leading wheelset (WS1) of the vehicle during the steady curve
resulted from the computer simulation as the most vulnerable wheelset to have wheel-climb derailment is the
leading wheelset. The associated derailment coefficients are shown in Figure 8. The results show that the forces and
the derailment coefficients calculated using analytic formulation are comparable to the results of the computer
simulation.

(a) (b)

FIGURE 7. Lateral (a) and vertical (b) forces at various speed, 400 m radius curve.
 FIGURE 8. Derailment coefficients at various speed, 400 m radius curve.

The evaluation was also conducted for the larger and smaller curve radius. Figure 9 and 10 exhibit the result for
600 meters curve radius while Figure 11 and 12 for 150 meters curve radius. For 600 m curving radius the result
also provides comparable values between analytic calculation and computer simulation. However, for the sharp
curve of 150 meters the significant different result is found. To explain this, we examined all of wheelsets position
when the vehicle negotiating the curve which the result is exhibited in Figure 11. This result is got from the
computer simulation.

FIGURE 9. Lateral (a) and vertical (b) forces at various speed, 600 m radius curve.

FIGURE 10. Derailment coefficients at various speed, 600 m radius curve.

 FIGURE 11. Lateral (a) and vertical (b) forces at various speed, 150 m radius curve.

FIGURE 12. Derailment coefficients at various speed, 150 m radius curve.

As shown in Figure 13 (a) and (b), for large and medium radius curves all wheelsets are move outward against
the outer rail and form the ideal position for negotiating the curve. For small radius curve of 150 m, as shown in
Figure 13 (c), the position of the leading wheelsets (WS1 and WS3) and the trailing wheelsets (WS2 and WS4) of
both bogies are quite different, where the leading wheelsets are move outward but the trailing wheelsets settle their
position in the center of the track, which is not the ideal position when passing the curve. As shown previously in
Figure 3, the unfavorable position of the trailing wheelset can produce anti-steering moment that negatively affects
the curving performance.
Because the turning forces work on the wheelset are much affected by the rolling radius difference between the
outer and the inner wheel and the flange contact, in such position of wheels shown in Figure 13 (c) the turning force
should mainly be taken by the leading wheelset. However, in analytic calculation we have assumed that the load of
the vehicle is distributed evenly to all the wheelset as well as all the forces working on it. That explains why the
result between the analytic calculation and the computer simulation at sharp curve are quite different. To
accommodate this, for the analytic procedure of calculation of derailment coefficient in the sharp curve, a correction
factor of 1.3 was applied into the formulation of flange force calculation in Equation 3, which represent the turning
force of the wheelset. The result for derailment coefficient after inclusion of the correction factor is presented in
Figure 14, which shows a good comparison to the results of the computer simulation.
 FIGURE 13. Wheelset position at curve: (a) R = 600 m (b) R= 400 m (c) R= 150 m radius.

FIGURE 14. Results after inclusion of correction factor in the analytic procedure

CONCLUSION
A simplified analytic procedure has been developed and a computer simulation using multibody dynamics
software Universal Mechanism has been performed to calculate the lateral and vertical forces at the contact point
between the wheel and the rail, which then used to calculate the derailment coefficient to evaluate the safety against
wheel-climb derailment when a railway vehicle negotiating the curved track. The results show that the values of
forces in the contact point between wheel and rail as well as the derailment coefficients calculated using simplified
analytical methods for large and medium curve are close enough to the values resulted from the computer
simulation. For the sharp curve a correction factor of 1.3 needs to be applied in the calculation of lateral force using
analytical procedure to match the result of the computer simulation. Based on the results from this study we can
conclude that the procedure of calculation using analytical method as described in this paper is proper to be used to
evaluate the derailment risk of the vehicle during the initial design of the new vehicle. However, the computer
simulation with full vehicle model and parameters still needs to be done in the later design stage to comprehensively
evaluate the derailment risk including in the transient zone where the assumption of quasi-static condition cannot be
applied.
 ACKNOWLEDGMENTS
Many thanks are addressed to PT Industri Kereta Api (Persero) for providing the vehicle design data needed for
computer simulation input and to Laboratory of Computational Mechanics, Bryansk State Technical University
Russia for allowing and supporting us to use Universal Mechanism, a multibody dynamics software platform to
analyze the dynamics of complex mechanical system.

REFERENCES
1. A.H. Wickens, Fundamental of Railway Vehicle Dynamics: Guidance and Stability (Swets & Zeitlinger B.V.,
Lisse, Netherlands, 2003).
2. S. Iwnicki, Handbook of Railway Vehicle Dynamics (CRC, Boca Rotan, FL, 2006).
3. X. Jun and Z. Qingyuan, “A study on mechanical mechanism of train derailment and preventive measures for
derailment” Vehicle System Dynamics, Vol. 43, No. 2, pp. 121-147 (2005)
4. Y. Nagumo, K. Tanifuji, J. Imai, “A Basic Study on Wheel Flange Climbing using Model Wheelset”
International Journal of Railway, Vol. 3, No. 2, pp. 60-67 (2010).
5. C.M. Kuo, C.C. Hiang Lin, “Analysis of derailment criteria”, Rail and Rapid Transit, pp.1-6, (2015).
6. C. Esveld, Modern Railway Track (MRT-Productions, Zaltbommel, Netherlands, 2001).
7. Kevin Oldknow, Wheel-rail interaction fundamental, (L.B. Foster Technology, 2015)
8. H. Takai, M. Uchida, H. Muramatsu, and H. Ishida, “Derailment Safety Evaluation Using Analytic Equations”,
Quarterly Report of RTRI, Vol. 43, pp. 119-124 (2002).
9. Universal Mechanism, “User’s Manual”, (Bryansk State Technical University, 2015).
10. W. Kik, J. Piotrowski, “A fast approximate method to calculate normal load at contact between wheel and rail
and creep forces during rolling”, Proceedings of 2nd Mini-Conference on Contact Mechanics and Wear of
Rail/Wheel Systems, pp. 338-340, (1996).
11. J. J. Kalker, “Wheel-rail Rolling Contact Theory”, Wear, pp. 243-261, (1991).

View publication stats