World Congress on Railway Research 2001, Köln, 25-29 November 2001

Derailment Safety Evaluation by Analytic Equations

Masao UCHIDA*, Hideyuki TAKAI*, Hironari MURAMATSU*, Hiroaki ISHIDA**

* Track Technology Division, ** Railway Dynamics Division
Railway Technical Research Institute, Japan

Summary
In estimating the risk of flange-climb derailment, it is generally considered that a wheel will
not derail if the derailment coefficient is smaller than the critical value calculated with
"Nadal's equation". To estimate derailment coefficients, time-series simulation is often done,
but this needs long calculation time. As a quicker method, we propose to estimate the derail-
ment coefficient from analytically calculated lateral and vertical wheel loads, base on analytic
equations and measured data. This paper describes the equations used, shows how the de-
railment coefficient is calculated, and compare with the results achieved by measurements.

Keywords: Flange-climb derailment, Vehicle dynamics simulation, Derailment coefficient,

Q/P estimation equations, Estimated Q/P ratio

1 Introduction
Recently, flange-climb derailments of light-weight commuter trains on tight curves at low
speed have tended to increase. This leads to a need for establishing safety evaluation methods
against flange-climb derailments.
Many factors influences the cause of derailments. In the field of vehicle eng ineering, these
are axle load, imbalance between the right and left-side static wheel loads, spring constants of
the primary and secondary suspensions, the height of the vehicle center-of -gravity and so on.
In the field of track engineering, curve radius, superelevation, twist, track irregularities and so
on. Beside these factors, the friction coefficients between the wheels and each rail in both the
tread and flange areas, and the train speed also play important roles.
To estimate the level of safety against flange-climb derailment, we traditionally use the "d e-
railment coefficient" known as "Nadal's equation". When the derailment coefficient is
smaller than the calculated value by Nadal's equation, we judge that there is no risk of flange-
climb derailment. On the other hand, in some cases, time-series computer simulations based
on vehicle dynamics model may be used to calculate the derailment coefficients for various
scenarios. However, because this method needs long calculation time, it is not suitable to
evaluate many cases that consist of combinations of numerous conditions.
We proposed equations that more easily estimate the wheel loads, lateral forces and d erail-
ment coefficients (lateral force/wheel load, usually described "Q/P" in Japan) from various
parameters. These equations consist of theoretical analysis of the mechanism of the lateral

1
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

force/wheel load on curves and of field test data. After estimating the lateral force and wheel
load of the outside rail, the derailment coefficient is obtained.
These equations are improved products of earlier equations that had been proposed u sing
numerous data from curving speed tests held in the 1990's. This method was effectively ap-
plied to investigate the cause of the unfortunate derailment accident that occurred on the Hi-
biya line of Teito Rapid Transit Authority (major subway operator in Tokyo) in March,
20001).
This paper expresses the Q/P estimation equations, calculation method of critical Q/P, e sti-
mation of safety margin against derailment using the "estimated Q/P ratio" (critical Q/P / es-
timated Q/P), and examples of trial calculations. This study aims at vehicles with bolster-less
bogies and air spring secondary suspension. The carbody is conservatively considered as a
rigid body.

2 Wheel load estimation equations

2.1 Wheel load variation by centrifugal force

When trains pass through curves, centrifugal forces act on the trains according to the curve
radius, superelevation and train speed. This adds to or subtract from the quasi-static outside
and inside wheel loads. If the train speed is lower than the balanced speed (= superelevation
excess), the wheel load of the outside wheel is less than its static value. On the contrary, if
the train speed is higher than the balanced speed (= superelevation deficiency), the wheel load
of the outside wheel will be higher than its static value. Considering the mechanism of in-
creased and decreased wheel loads by the effect of centrifugal forces, equations (1) and (2)
are formed to estimate the quasi-static outside and inside wheel loads.

W0 v2 C H G?– v 2 C
Po ?E
2 gR G....................................................................................................
G 2 gR G (1)

?–
W0 v2 C HG v2 C
Pi 2 ..............................................................................................
?E (2)
2 gR G G 2 gR G

Here,
Po : constant component of wheel load of the outside wheel (kN)
Pi : constant component of inside wheel load (kN)
_: static wheel load ratio of outside wheel
W0: static axle-load (kN)
v: train speed (m/s)
G: gauge (m)
C: track cant (m)

2
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

R: curve radius (m)

g: gravity acceleration (9.8 m/s2)
HG*: effective height of center-of-gravity of vehicle (m)
(For non-tilt vehicles regardless of anti-roll bars, HG* is set to 1.25 times of the
real height of the center-of-gravity)

2.2 Wheel load variation due to track twist

On transition curves, as the track surface is twisted, wheel loads increase and decrease a c-
cording to the deformation of the primary and secondary suspension springs. Especially, on
the exit transition curve, the wheel load of the outside wheel of the leading axle decreases ac-
cording to the extension of its primary suspension. At the same time, the wheel loads of the
outside wheel of the whole leading bogie decreases according to the extension of the secon-
dary suspension spring. Furthermore, the wheel loads vary by local track irregularities. The
static wheel load decrease is expressed by equation (3), considering the mechanism of the
wheel load variation due to track twist.

1 tc
P ?E
K t a?Ek1
8b 2 2
1 2
K ...............................................................................................
k 1 2k1 b1 (3)
1 1
2 2
4k 1b1 2k 2 b2

2c 2a
tc tc ta ta
aTC a TC

Here,
_Pst: Static wheel load decrease due to track twist (kN)
K_: Effective rotational stiffness / wheelset (kN-mm)
K' _1: Effective rotational stiffness / bogie (kN-mm)
2b: Width between right and left wheel/rail contact points (mm)
2b1: Width between right and left primary suspension springs (mm)
2b2: Width between right and left secondary suspension springs (mm)
k1: Primary suspension vertical stiffness / axlebox (MN/m)
k2: Secondary suspension vertical stiffness / bogie side (MN/m)
_tc : Track twist between bogie centers (mm)
_ta: Track twist between a wheelsets (within bogie) (mm)
2a: Distance between wheelsets (within bogie) (mm)
2c: Distance between bogie centers (mm)
aTC: Cant gradient
tc : Track twist between bogie centers excess cant gradient (mm)

3
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

ta: Track twist between wheelbase excess cant gradient (mm)

2.3 Wheel load variation due to torsion of secondary suspension spring

When passing through curves, torsion of secondary suspension springs will occur on a bo l-
ster-less bogie due to the relative rotational deformation between carbody and bogies. The re-
action force F1 acts laterally on the rail. At the same time, F1' (= F1/tan60°), the vertical com-
ponent of the reaction force, will act vertically (Figure 1). Track shifting force F1 that occurs
by the torsion of air springs is described in Section 3.2, Clauses (2) and (3).

2.4 Estimation equations for of wheel loads

The outside and inside wheel loads can be calculated by the equation (4) and (5), consi der-
ing three factors: centrifugal force, track twist and torsion secondary suspention springs.

P0 P0 ...................................................................................................................
P F1 (4)
tan 60
Pi Pi ...................................................................................................................
P F1 (5)
tan 60

Here,
Po: Wheel load of the outside wheel (kN)

4
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

Pi: Inside wheel load (kN)

Po : Constant component of wheel load of the outside wheel (kN)
Pi : Constant component of inside wheel load (kN)
P : Static wheel load decrease by track twist (kN)
F1: Track shifting force by distortion of air springs (kN)
_: Collection coefficient of vertical component of F1 (See chapter 3.2)

3. Lateral force estimation equations

3.1 Turning lateral force due to reaction of the inside friction force
When a vehicle is running on a curve, the flange of the leading, outside wheel is in contact
with the outside rail and is pushed against its gauge face. Then the inside wheel resists the
force with a friction force (= product of wheel load and friction coefficient applied on the
tread). This acts as a quasi-static lateral force toward the outside, that is "turning lateral
force" (Figure 2). Accordingly, the larger the friction coefficient between the inside wheel
tread and the rail (nearly inside Q/P ratio _) is, the larger the turning lateral force becomes.
The estimation equation for the inside wheel quasi-static component of the lateral force
(turning lateral force) is expressed by equation (6).

Qi = _Pi ..................................................................................................................... (6)

Here,
Qi : Constant component of inside lateral force (kN)
_: Inside Q/P ratio
Pi: Inside wheel load (kN)

Two values for _ are established for tapered and arc or modified arc wheel profiles, r espec-
tively (Figure 3). These characteristics are obtained by time-series simulations and field test
data.

5
World Congress on Railway Research 2001, Köln, 25-29 November 2001

6
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

3.2 Track shifting force due to centrifugal force and torsion of secondary suspension
springs
(1) Track shifting force due to centrifugal force
Centrifugal forces act on the train that is running through curves according to the curve r a-
dius, superelevation and train speed. It constitutes one part of the quasi-static track shifting
force. This force is negative when the train speed is lower than the balanced speed (superel-
vation excess). On the contrary, this force is positive when the train speed is higher than the
balanced speed (superelvation deficiency).
(2) Track shifting force due to torsion of air spring
On curves, the torsion of the secondary suspension springs due to the yaw angle b etween
carbody and bogies cause a track shifting force. At the position of the leading wheelset of a
bogie, this force acts towards the outside rail as shown in Figure 4.

(3) Estimation equations for quasi-static component of track shifting force

Based on the above mentioned (1) and (2), quasi-static component of the track shifting force
due to the centrifugal force and torsion of the secondary suspension springs is expressed by
the following equation (7).
v2 C
QAS W0 F1
gR G

7
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

v2 C 2kb22 c
W0 10 6
................................................................................................ (7)
gR G aR

Here,
_QAS: Quasi-static component of track shifting force (kN)
F1: Track-shifting force due to deformation of secondary suspension springs (kN)
W0: Static axle load (kN)
v: Train speed (m/s)
G: Gauge (m)
C: Superelevation (m)
R: Curve radius (m)
g: Gravitational acceleration (9.80 m/s2)
2b2: Width between right and left air springs (mm)
2a: Distance between wheelsets (within bogie) (mm)
2c: Distance between bogie centers (m)
k: Yaw stiffness of secondary suspension spring / bogie (kN/m)
(zero for bogies with bolster)
_: Modifying coefficient of track shifting force F1

(4) Modifying coefficient _ of the track shifting force due to torsion of secondary suspension
springs
According to the result of study on the calculation value of track shifting force F1 due to the
deformation of air springs by estimation equation and, time series simulation (Figure 5), we
set modifying coefficient _ per equations (8) and (9)

(a) In case of inside Q/P ratio _ 0.50

0.7 ?@?@?@?@?@?@
?@
R 160
310 R
0. 7 .....................................................................................................
?@ 160 R 1000 (8)
150
3.2 ?@ ?@?@ 1000 R ?@?@?@?@?@?@?@

(b) In case of inside Q/P ratio _ > 0.50

0.7 ?@?@?@?@?@?@
?@
R 160
310 R......................................................................................................... (9)
0.7 ?@160 R 1000
150
3 .2 ?@ ?@?@ 1000 R ?@?@?@?@?@?@?@

8
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

In Figure 5, when the curve radius is larger than a certain value, the track shifting force F1 due
to the torsion of the secondary suspension springs becomes negative. Consequently, for the
outside wheel, the reaction force of the vertical component of F1 acts downward. This force
reduces the wheel load of the outside wheel. However, such a condition was not found during
field tests nor by time series simulation. Therefore, the modifying coefficient for vertical
component of F1 is set per equation (10).

_ = 1 ( _ > 0 ), _ = 0 ( _ 0 ) .................................................................................. (10)

3.3 Lateral force variation due to track irregularities and impacts at rail joints
When there are track irregularities, especially alignment irregularities, variation of track
shifting forces occur mainly by inertia forces according to the vehicle vibrations. Shocking
variation of lateral forces at rail joints occur. These forces increase with higher train speed.
The way of calculations are shown in equations (11) and (12).

Q AD 3W0 k.............................................................................................................
Q ZV (11)

2500 V
Qunsp 4
............................................................................................................. (12)
100 R 100

Here,

9
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

_QAD: Variation of track shifting force (kN)

W0: Static axle load (kN)
_z: Standard deviation of alignment irregularities (mm)
V: Train speed (km/h)
kQ: Variation coefficient of track shifting force (1/mm/(km/h))
_Qunsp: Variation of lateral force at rail joints (kN)
R: Curve radius (m)
_: Effective ratio of variation of lateral force (%)

3.4 Estimation equations of outside lateral forces

We estimate the outside lateral forces using equation (13) considering the above mentioned
three factors: turning lateral forces by the inside rail friction force, track shifting force due to
centrifugal force and torsion of secondary suspension springs, and variation of lateral forces
due to track irregularities and impact forces at rail joints.

Qo Qi QAS ........................................................................................................
QAD Qunsp (13)

Here,
Qo: Outside lateral force (kN)
Qi : Constant component of inside lateral force (kN)
_QAS: Constant component of track shifting force (kN)
_QAD: Variable component of track shifting force by track irregularity (kN)
_Qunsp: Variable component of lateral force at rail joints (kN)

4 Calculation of critical derailment coefficients

4.1 Critical derailment coefficients

We usually use the "Nadal's equation" to calculate the critical derailment coefficient. Here,
instead of the friction coefficient _, we adopt the equivalent friction coefficient _e which is a
function of the wheel angle of attack, . This is to reflect the difference of the track geometry
(curvature) accurately at the wheelset under consideration. This process gives higher critical
derailment coefficients than with the common friction coefficient _.

4.2 Nadal's equation

Nadal's equation, which is used for calculating the margin against the flange-climb d erail-
ment quantitatively, is shown as equation (14).

tan e
Q P cri
1 e tan
10
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

............................................................................................................... (14)

Here,
(Q/P)cri: Critical derailment coefficient
_: Wheel flange angle (rad)
_e : Equivalent friction coefficient = fy /N (fy : lateral creep force, N: normal force)

4.3 Approximation of equivalent friction coefficient _e

The equivalent friction coefficient _e is expressed as equation (15), considering the satur a-
tion characteristics of the creep force.

22 y N
e .............................................................................................................
1 (15)
22 y N
Here,
_: Index expressing saturation characteristic ( = 1.5)
vy : Lateral creep ratio
_22: Lateral creep coefficient
N: Normal force
_22/N 27.0

Considering that v = tan_ · sin_ (1–cot2_) _ _ ( _ is the wheel angle of attack), equ ation
(16) is obtained.

27.0
e ..........................................................................................................
2 (16)
1 .5 1 .5
27.0 3

Here, _ is the effective value of friction coefficient at the outside wheel flange with co nsid-
ering the forward/rear direction of the tangent force.

4.4 Setting method of angles of attack

For vehicles with tapered, arc or modified-arc wheel tread, considering the result of anal ysis
of time-series simulation (Figure 6), the wheel angle of attack _ is calculated by equation (17).

a .................................................................................................................... (17)
?? T w
R
(a) For tapered wheel tread
90 1 2
T W ?@?@?@
R 90
R 2a

11
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

1 .......................................................................................................
2 (18)
T W ?@?@ ?@?@ R 90
2a
(b) For modified arc wheel tread

60 1 2
T W ?@?@?@
R 200
R 2a

1 2 0 .7 1 1
T W 0 .3 ........................................................................................
?@?@ 80 ?…R?ƒ200 (19)
2a 0.0075 R 200

1 2
T W ?@?@
?@?@
R?ƒ80
2a

Here,
_: Attack angle (rad)
2a: Wheelbase (m)
R: Curve radius at the center of leading bogie (m)
_T: Bogie yaw angle (including carbody yaw angle) (rad)
_W: Yaw angle by the bogie staring (rad)
_1: Side gap + gauge widening/2 at leading axle (m)
_2: Side gap + gauge widening/2 at trailing axle (m)

12
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

5 Evaluation of the margins to the flange-climb derailments using esti-

mated derailment coefficient ratio

5.1 Estimation of outside derailment coefficients

The derailment coefficient of the outside wheel is calculated by equation (20) as the ratio of
outside lateral force to wheel load of the outside wheel at the same position.

Q
)o?? o
(Q / P....................................................................................................................... (20)
Po

Here,
(Q/P)o: Derailment coefficient
Qo: Outside lateral force (kN)
Po: Wheel load of the outside wheel (kN)

5.2 Estimated derailment coefficients ratio

We definite an "Estimated derailment coefficient ratio" as the ratio of the critical Q/P shown
in Chapter 4 to the Q/P of the outside wheel shown in Section 5.1. This definition is shown in
equation (21). This ratio expresses the margin against flange-climb derailment with the base
value of 1.0.

Q P cri
??
(Q / P )ratio.................................................................................................................. (21)
Q Po

Here
(Q/P)cri: Critical derailment coefficient
(Q/P)o: Outside derailment coefficient

5.3 Trial calculation of estimated derailment coefficient ratio (Sensitivity analysis)

(1) Comparison of Q/P between estimation and measurement
Figure 7 shows result of comparison of Q/P between estimation and measurement. E sti-
mated values are the good approximations of measured data.
(2) Example of estimated Q/P ratio
Figure 8 shows an example of calculation result at the section of a circular curve and the
exit transition curve, with a calculation interval of 1 meter. The estimated Q/P ratio has a
minimum just after entering the exit-side transition curve.
(3) Sensitivity analysis of estimated Q/P ratio
We calculate the estimated Q/P ratio in various cases with changing p arameters shown in
Table 1 as basic values. Figure 9 shows the calculation result. Here, those values are mini-
mum value in the calculation area.

13
World Congress on Railway Research 2001, Köln, 25-29 November 2001

14
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

Calculation parameters

15
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

6 Conclusions
(1) We composed the wheel load estimation equation considering centrifugal forces, track ge-
ometry deviation and the deformation of secondary suspension springs.
(2) We established the lateral force estimation equation considering the curve turning lateral
force, track shifting force due to centrifugal force and secondary suspension springs, varia-
tion of lateral force due to track irregularities.
(3) We proposed a calculation method for the critical derailment coefficient considering
wheel flange angle, equivalent friction coefficient and wheel angle of attack.
(4) We define the "Estimated derailment coefficient ratio" as the ratio of critical Q/P to the
outside Q/P, and evaluate the margin against the flange-climb derailment.

In this report, some parts, such as establishing method for the inside wheel Q/P ratio or
modification of track shifting force, were set up provisionally. Hereafter it is necessary to
improve the accuracy according to theoretical study and field data. At this stage, this method
is only for bolster-less bogies with air spring secondary suspension. We intend to improve the

16
 World Congress on Railway Research 2001, Köln, 25-29 November 2001

equations to apply also to other types of bogies.

References
1) Railway accident investigation committee, Ministry of Transportation : Investigation Re-
port of Train Derailment at Naka-Meguro Station on the Hibiya Line of the Teito Rapid Tran-
sit Authority (in Japanese), 2000.10

17