Engineering Failure Analysis 116 (2020) 104725

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Fatigue life assessment methods for railway vehicle bogie frames

T
⁎
Ruixian Xiua,b, , Maksym Spiryaginb, Qing Wub, Shuchen Yanga, Yanwen Liuc
a
College of Engineering, Changchun Normal University, Jilin, China
b
Centre for Railway Engineering, Central Queensland University, Rockhampton, Australia
c
Engineering Laboratory, Centre of Rolling Stock Engineering Research Changchun Railway Vehicles Co., Ltd., Jilin, China

A R T IC LE I N F O ABS TRA CT

Keywords: The fatigue life of bogie frames directly determines railway vehicles’ reliability and safety.
Bogie frame Assessing bogie frame fatigue life is a significant requirement for railway vehicle designs. This
Fatigue life assessment method paper reviews current railway bogie frame fatigue life assessment methods. These methods use
Load spectrum standard-prescribed loads, measured load spectrums or simulated load spectrums to assess fa-
tigue life. The methods of using standard-prescribed loads are categorised into design load
methods; they are conservative and are usually used in the design stage or for remaining life
assessment. The methods of using measured or simulated load spectrums are categorised into
time domain method, frequency domain method, time-frequency domain method, Road
Environment Percent Occurrence Spectrum (REPOS) method and bench test method; their as-
sessment has a good agreement with bogie frames’ real fatigue life, but they cannot be used for
assessing crack propagation. As for bogie frame fatigue life assessment, standard-prescribed loads
and measured load spectrums take into account traction and braking loads. But design loads
prescribed by standards cannot reproduce operational scenarios; measured load spectrums re-
produce operational scenarios but require longer simulation or measured times, higher cost and
are not suitable for the design stage. One research gap has been identified which is that current
fatigue life assessment methods based on simulated load spectrums do not consider traction and
braking forces.

1. Introduction

Due to the rapid development of the global economy, passenger and freight traffic has increased sharply, causing transportation
systems to face huge challenges in capacity and speed. Compared with road and air transportation, rail transportation has advantages
in higher capacity, lower cost, lower emissions and lower energy consumption. To satisfy current transportation requirements as well
as enhance rail transportation’s merits, high-speed and heavy-haul railway vehicles are widely used for main line operations.
To achieve high-speed and heavy-haul, powered railway vehicles are required to have higher power. At the same time, railway
vehicle components should be designed as lightweight to reduce energy consumption, and to maximise speed and payload. But in
high-speed and heavy-haul operational scenarios, railway vehicle components are subjected to severe vibration loads. This may make
railway vehicles’ fatigue problems more serious, which would directly affect their operational safety. As the key load-bearing
component of any railway vehicle, a bogie frame is responsible for supporting the car-body and other necessary components. A bogie
frame’s fatigue life determines a railway vehicle’s reliability and safety. Hence, it is a significant requirement to carry out bogie frame
fatigue life assessments, especially in high-speed and heavy-haul operational scenarios.
This paper first reviews standard-prescribed loads applied for fatigue life assessment. Then, measured and simulated load

⁎
Corresponding author.

https://doi.org/10.1016/j.engfailanal.2020.104725
Received 2 September 2019; Received in revised form 29 June 2020; Accepted 2 July 2020
Available online 09 July 2020
1350-6307/ © 2020 Elsevier Ltd. All rights reserved.
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

spectrums that reproduce the operational scenarios are discussed. This paper also reviews how to use standard-prescribed loads and
measured or simulated load spectrums to assess bogie frame fatigue life.

2. Standard-prescribed loads

This section reviews standard-prescribed loads used for railway vehicle bogie frame fatigue life assessments. The standards
reviewed include JIS 4207 [1], UIC 515-4 [2], UIC 615-4 [3] and EN 13749 [4]. These standards are applied for railway vehicle bogie
frames (excluding specific railway vehicles equipped with rubber tires). Standard-prescribed loads are categorised into design loads.
Design load is the load that be attributed a degree of standardization for pseudo-static (so-called static) and cycling loads. Load cases
are commonly vary depending on the type of testing procedures as defined in the relevant standards and those cases are not con-
nected with actual operational cases.

2.1. Design (static) load

The JIS 4207 [1], UIC 515-4 [2], UIC 615-4 [3] and EN 13749 [4] standards prescribe the predefined loads for the static strength
test used for fatigue life assessment; these loads have constant values. For JIS 4207 [1], loads of the static strength test used for
fatigue life assessment are divided into static and dynamic loads. Static loads refer to the load-bearing masses (e.g., bogie frame’s
mass, car body’s mass, payload and bogie frame components’ masses) acting on a bogie frame when the vehicle is in the static state,
see Eq. (1).

W = W1 + W2 + W3 (1)

Dynamic loads are affected by many factors, including the vibration of the car body and bogie, track irregularity and traction
running conditions, etc. JIS 4207 [1] prescribes various loads that railway vehicles are subjected to in service, which especially
considers in detail the loads generated by mounted components such as traction and braking loads. However, as track irregularity,
speed and the frequency of dynamic load occurrences vary in different situations, most dynamic loads are given in ranges as shown in
Table 1. The coefficients of dynamic loads are determined as empirical values; they are used to represent those loads that occur
frequently in service. Twist loads do not consider the effect of the bogie frame’s wheelbase. Note that all dynamic loads are variable
loads acting on the bogie frame in normal service. JIS 4207 [1] does not prescribe exceptional loads. Another point worth mentioning
is that these dynamic loads (e.g., traction and braking loads) are applied in the static state, and therefore cannot reproduce actual
vibration loads in operational scenarios.
Different from JIS 4207 [1], for UIC 515-4 [2], UIC 615-4 [3] and EN 13749 [4] standards, loads of the static strength test used for
fatigue life assessment are divided into normal loads, exceptional loads and particular in-service loads, which are calculated by using
specific formulas as shown in Table 2. For a 3-axle bogie frame, since the centre wheel-set does not participate in lateral load
transmission [3], the number of axles ne is 2 in the calculation of lateral loads. Note that most loads have different values in these
standards. For example, UIC 515-4 [2] specifies a damper’s loading force as 1.2 times the damper’s unloading force; UIC 615-4 [3]
specifies a damper’s loading force as 1.5 times the damper’s unloading force; EN 13749 [4] specifies a damper’s loading force as equal
to the damper’s unloading force. Another point that should be noted is that the longitudinal loads in UIC [2,3] are regarded as
particular in-service loads. UIC 615-4 [3] specifies the applied location of traction loads; and the traction loads’ values are decided by
the designer for each considered vehicle. EN 13749 [4] prescribes that loads resulting from braking or traction motors are categorised
into exceptional loads and normal service loads. In all these standards, twist loads are usually described by the track twist dis-
placement. For UIC [2,3] and EN [4], twist loads are related to the bogie frame’s wheelbase, see Eq. (2) [5].

h = la × ×i
lp
lr (2)

UIC 515-4 [2], UIC 615-4 [3] and EN 13749 [4] standards based on prescribed loads (see Table 2) propose 17 main in-service load
cases as shown in Table 3. These load cases are the combinations of main loads (e.g., vertical loads, lateral loads and twist loads)
encountered during straight and curve negotiations, to verify the risk of a bogie frame’s fatigue cracking. These load cases are also
used to evaluate the effects of lateral and twist loads on bogie frame fatigue life. Note that quasi-static and dynamic variations of
vertical loads induced by car body rolling and bouncing are considered in the main load cases. Dynamic vertical loads are determined
by using a rolling coefficientα and a bouncing coefficient β . For powered bogie frames, these load cases also include traction motor
loads; how to apply these loads in main load cases depend on the actual operational scenarios. Loads acting on the bogie frame are
shown in Fig. 1 and Fig. 2, where lateral loads are applied to lateral bumper stops and springs.
For UIC 515-4 [2], UIC 615-4 [3] and EN 13749 [4] standards, only EN 13749 [4] specifies longitudinal load cases (as shown in
Table 4) and exceptional load cases in detail. Longitudinal load cases are the combination of vertical loads and longitudinal loads.
Since the damage to bogie frames is mainly fatigue damage, exceptional load cases are not as many as the main in-service load cases
which only include several large main loads, such as vertical loads, lateral loads and twist loads. Exceptional load cases are defined as
follows: all exceptional loads and combinations of exceptional loads that occur in service; the combinations of vertical loads (cor-
responding to an empty vehicle) and exceptional twist loads; exceptional vertical loads and particular in-service loads applied se-
parately or in combination.

2
 R. Xiu, et al.

Table 1
JIS dynamic loads of static test [1].
Type Load condition Load values Applicable (example)

Vertical dynamic loads Vertical load caused by load-bearing masses’ vibration

Vertical dynamic load caused by mounted Side beam mounted Braking component
W21: (0.2 − 0.5)W

components masses’ vibration components

W22: (1 − 2) LP

Cross beam mounted Traction motor and driving

components component
W23: (3 − 10) LP

End beam mounted Braking component and

components cowcatcher
W24: (5 − 10) LP

Driving load acting on the driving seat Gear box hanger and reducer
bracket
W25: (0.2 − 0.4)P

Braking load acting on the braking seat Braking unit and braking disc
Vertical damper load Shaft spring damper
W26: P × f
W27:determined by damper characteristics
Anti-yaw load W28:determined by anti-yaw component characteristics Anti-yaw bearing component

3
Lateral dynamic loads Lateral dynamic load caused by load-bearing masses’ vibration and centrifugal
force
WY 1: (0.2 − 0.4) W

Lateral bumper stop load WY 2: WY1minus the loads acting on the secondary spring Lateral bumper stop
Lateral dynamic load caused by mounted components masses’ vibration Traction motor and braking
component
WY 3: (2 − 4) LP

Lateral damper load WY4:determined by damper characteristics Lateral damper

Longitudinal dynamic loads Longitudinal load caused by load-bearing masses’ vibration

Traction load
WX1: (0.2 − 0.4)W

Longitudinal load caused by mounted components masses’ vibration Traction motor and braking
WX 2: (0.2 − 0.4)P

component
WX 3: (1 − 3) LP

Braking load WX 4: P

Twist loads Twist displacement by super high reduction Equal to the loads that the plane deformation displacement 10–15 mm of
the track at the diagonal position of the wheel
Engineering Failure Analysis 116 (2020) 104725
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Table 2
UIC/EN loads of static strength test in-service conditions [2–4].
Types of loads Normal loads Exceptional loads Particular in-service loads

Fz = + C1)− nb m+⎤(EN: light Fzmax = + C2) − nb m+⎤(EN: light rail Damper load: 1.2FA (UIC 515 − 4)
⎡ (mv
nb ⎣
⎡ (mv
⎦ nb ⎣ ⎦ 1.5FA (UIC 615 − 4) FA (EN 13749)
Vertical loads Fz g c 1.4g c
(per bogie) 100 100

Fz = + 1.2C1 − nb m+)
rail vehicles and trams) vehicles and trams)
Fzmax = + C2 − nb m+) (EN:
g 1.4g
(mv (mv if operating
nb nb
conditions are very poor, the factor 1.4 may be

Fy = 0.5(Fz + 0.5m+g ) (UIC)

increased up to 2.0.)
⎛ (mv + C2 ) ×g⎞
Fymax = 2 ⎜10 4 + FB = fanti − roll × α
Lateral loads Fy c Anti-roll bar load FB :
Fy = (Fz + m+g )/4 (EN) ⎟ (EN: light rail
100

⎝ ⎠
(per bogie) 3ne nb

vehicles and trams)

Fymax = 2 ⎛10 4 + v 2 ⎞
(m + C ) g
⎝ 3ne nb ⎠

Fx1 = 0.05(Fz + m+g ) (EN: due to dynamics) Fx1max = 0.1(Fzmax + m+g ) (EN) Fx1 = 0.05(Fz + m+g ) (UIC)
Fx1 = 0.1(Fz + m+g ) (EN: due to tractive
Longitudinal
lozenging
loads Fx1(per effort)
wheel)
Twist loads Track twist of 0.5% Track twist of 1% (UIC/EN) Inertia loads of traction motors and
Ft1, Ft2 Or vehicle empty under vertical load consider transmission massesFC
complete unloading of one wheel with the vertical
displacement of the wheel being limited to rail height
(EN)
Traction loadsFD 1.1 times the normal loads induced during 1.3 times those produced during starting or dynamic FD (UIC 615)
normal service starting or stopping (EN) braking with the maximum acceleration or
deceleration (EN)
Braking loadsFb 1.1 times the nominal forces induced by 1.3 times those produced during emergency braking Fb (UIC 615)
the service braking (EN) (EN)

Table 3
UIC/EN main load cases of static test [2–4].
Load case Vertical load Lateral load Twist load

Fz1 Fz2 Fy Ft1 Ft2

0 Installation of traction motors

(UIC 515-4 doesn’t have traction motors’ forces)

(1 + α − β ) Fz /2 (1 − α − β ) Fz /2
1 Fz/2 Fz/2

(1 + α − β ) Fz /2 (1 − α − β ) Fz /2 + Fy
2

(1 + α + β ) Fz /2 (1 − α + β ) Fz /2
3

(1 + α + β ) Fz /2 (1 − α + β ) Fz /2 + Fy
4

(1 − α − β ) Fz /2 (1 + α − β ) Fz /2
5

(1 − α − β ) Fz /2 (1 + α − β ) Fz /2 − Fy
6

(1 − α + β ) Fz /2 (1 + α + β ) Fz /2
7

(1 − α + β ) Fz /2 (1 + α + β ) Fz /2 − Fy
8 0

(1 + α − β ) Fz /2 (1 − α − β ) Fz /2 +
9

(1 + α − β ) Fz /2 (1 − α − β ) Fz /2 +
10 Fy Ft1

+
Fy Ft2
(1 + α + β ) Fz /2 (1 − α + β ) Fz /2
11

+
Fy Ft1
(1 + α + β ) Fz /2 (1 − α + β ) Fz /2
12

−
Fy Ft2
(1 − α − β ) Fz /2 (1 + α − β ) Fz /2
13

−
Fy Ft1
(1 − α − β ) Fz /2 (1 + α − β ) Fz /2
14

−
Fy Ft2
(1 − α + β ) Fz /2 (1 + α + β ) Fz /2
15

−
Fy Ft1
(1 − α + β ) Fz /2 (1 + α + β ) Fz /2
16
17 Fy Ft2

2.2. Design (cyclic) load

It is known that fatigue is a mechanism that results from cyclic loads. So, research on bogie frame fatigue life should consider the
effects of cyclic loads. JIS 4207 [1] does not specify cyclic loads for fatigue tests. However, UIC 515-4 [2], UIC 615-4 [3] and EN
13749 [4] have made provisions for cyclic loads for bogie frames’ fatigue tests.
For UIC 515-4 [2], UIC 615-4 [3] and EN 13749 [4] standards, the main loads of the fatigue test include vertical loads (static,
quasi-static and dynamic vertical loads), lateral loads (quasi-static and dynamic lateral loads) and twist loads as shown in Table 5.
Quasi-static components of vertical loads take into account the rolling motion of the car-body during curve negotiations. Dynamic

4
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 1. Bogie frame loading arrangement for static test [2–4].

Fig. 2. Schemes of load applications for bogie frame fatigue analysis: (a) 2-axle bogie frame, (b) 3-axle bogie frame.

Table 4
EN longitudinal load cases [4].
Load case Fz1 Fz2 Fx

+Fx1
1 Fz/2 Fz/2 0

−Fx1
2 Fz/2 Fz/2
3 Fz/2 Fz/2

Table 5
UIC/EN loads of fatigue test [2–4].
Types of loads Vertical loads Lateral loads Twist loads

Fzs1 = Fzs2 = Fz
Fzq1 = −Fzq2 = ±αFz Fyq = ±0.25(Fz + 0.5m+g ) (UIC [2,3])
Static component
Quasi-static component
Fyq = ±0.063(Fz + m+g ) × ne (EN [4])
Dynamic component Fzd1 = −Fzd2 = ±βFz Fyd=±0.25(Fz + 0.5m+g ) (UIC [2,3])
Fyd = ±0.063(Fz + m+g ) × ne (EN [4])
Track twist of 0.5%,
Ft1, Ft2

components of vertical loads take into account the vertical motion (bouncing) of the car-body during curve negotiations. Lateral loads
come from track irregularity and inertial forces caused by curve negotiations. Quasi-static components of lateral loads are non-
compensated forces. Dynamic components of lateral loads result from track irregularity. The quasi-static components and dynamic

5
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Table 6
Bogie frame dynamic loads of each stage for fatigue test [2–4].
Loading sequences Lateral loads Vertical loads Number of cycles

Fy = Fyq + Fyd Fz1 = 0.5(Fzs + Fzsq + Fzd ) Fz2 = Fz1 6 × 106

Fy = 1.2 × (Fyq + Fyd ) Fz2 = Fz1
The first stage
Fz1 = 0.5(Fzs + 1.2Fzsq + 1.2Fzd ) 2 × 106
Fy = 1.4 × (Fyq + Fyd ) Fz2 = Fz1
The second stage
The third stage Fz1 = 0.5(Fzs + 1.4Fzsq + 1.4Fzd ) 2 × 106

components of lateral loads prescribed in UIC [2,3] and EN [4] standards are described by various equations, see Table 5. All these
standards do not prescribe additional loads. But they point out that, if necessary, additional loads (e.g., traction loads, braking loads
and damper loads) can be used to assess bogie frame fatigue life, which obviously depends on actual service conditions.
For UIC 515-4 [2], UIC 615-4 [3] and EN 13749 [4] standards, the cyclic loads (see Table 5) used for fatigue test are applied in
three stages as shown in Table 6. EN 13749 [4] prescribes that the loads used for the fatigue test are vertical loads, lateral loads and
twist loads. For UIC 515-4 [2] and UIC 615-4 [3], twist loads exist in specific load cases. When the static strength test shows that twist
loads have nothing to do with bogie frames, there is no need to consider twist loads in the fatigue test. When the static strength test
indicates that the results of applied twist loads are different from applied vertical and lateral loads, twist loads and the combinations
of vertical loads and lateral loads should be applied to each bogie frame separately in the fatigue test. Otherwise, twist loads, vertical

and dynamic load coefficients in each stage. The total number of cycles of twist loads is 106 ; application of the twist loads is 0.6 × 106
loads and lateral loads can be applied together in the fatigue test. And the coefficient of a twist load corresponds to the quasi-static

cycles in the first stage; application of the twist loads is 0.2 × 106 in the other two stages, respectively. And twist loads are applied
after the end of the application of vertical loads and lateral loads in each stage. The dynamic cyclic loading arrangement is shown in
Fig. 3. Load application locations of the fatigue test are shown in Fig. 1.

3. Measured load spectrums

This section reviews measured load spectrums of bogie frames derived from field tests; these load spectrums can be used for bogie
frame fatigue life assessments. Research papers indicate that measured load spectrums are divided into the REPOS load spectrum,
dynamic stress test load spectrum and Power Spectrum Density (PSD) load spectrum.

3.1. REPOS load spectrum

As early as 1980, AAR [6] conducted a series of tests for the coupler, centre plate and side bearer loads in operational scenarios.
Researchers proposed a methodology to use REPOS for the fatigue assessment of freight cars.
REPOS derived from field measured data under appropriate samples, characterising the structural loading experienced during the
total service of life and reflecting real operational scenarios. Strain gauges are installed on test cars at locations which should have
high sensitivity for the intended load measurement but relatively low sensitivity for the other applied loads [6]. Optimum locations
for strain gauges can be determined by using FE analysis.
REPOS is a three-dimensional load spectrum using the maximum load and minimum load as its coordinates. For each maximum-
minimum load combination, the third dimension entered is the percentage of occurrences with this load combination encountered in
the environment. As shown in Fig. 4, REPOS is a folding spectrum. All the percentage values are in the upper left diagonal location.
The reason for this is the Palmgren-Miner rule which assumes the order of load application does not affect fatigue damage [7].
References [6] and [8] introduce the process for the development of REPOS and the fatigue life assessment method. Baker et al.
[9] illustrated the process for obtaining the Australian REPOS. The rain flow counting technique is used to convert measured data into
REPOS [8]. Note that the rain flow counting technique [10] is the most popular and successful way to extract the fatigue content from
a load-time history. It has the ability of the algorithm to count full cycles and match similar half-cycles into pairs, which form
complete cycles [11,12].
Using REPOS to assess bogie frame fatigue life is effective, quick and economical. How to convert a load spectrum (REPOS of
force) into a stress spectrum (REPOS of stress) is significant for fatigue life assessment. An alternative approach that involves direct
measurement of the stress on the critical locations of bogie frames is used to obtain a stress spectrum [8] whereby a dynamic stress

Fig. 3. Dynamic cyclic loading.

6
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 4. An example of REPOS.

test is performed to get the stress-time history, which is converted directly to the stress spectrum for bogie frame fatigue life as-
sessments.

3.2. Dynamic stress test load spectrum

A dynamic stress test in operational scenarios is the most direct and effective way to obtain the bogie frames’ load spectrum (stress
spectrum) [13]. Strain gauges are fixed at the critical locations of the bogie frames to obtain its stress-time history via a data
collection system. With the help of the rain flow counting technique, the stress-time history is converted into stress spectrum [14–16].
The most simplified stress spectrum is a two-dimensional stress spectrum. It regards the mean stress as a constant, and the stress
amplitude or range is treated as a random variable. The two-dimensional stress spectrum consists of the stress amplitude (or stress
range) and the number of the stress amplitude's cycles encountered in the environment. Luo et al. [17] used the stress range and the
number of stress range's cycles to represent a bogie frame’s two-dimensional stress spectrum.
But for some bogie frames, the mean stress value varies greatly. When mean stress is assumed as constant, it would inevitably
result in some inaccuracy [18]. Therefore, Baek et al. [19], Wu et al. [20], Li et al. [21] and Ma et al. [22] treat mean stress and stress
amplitude as random variables. The random variables are combined with the number of corresponding load cycles to produce a three-
dimensional load spectrum as shown in Fig. 5. A three-dimensional load spectrum provides a more accurate assessment for bogie
frame fatigue life. However, the disadvantage of the dynamic stress test is that sufficient test data must be obtained for truly
representative stress spectra. This is time-consuming and costly, which may result in the situation that only part of the railway
vehicles can be selected for the dynamic stress test. Since the dynamic stress test is performed after the bogie frame has been
produced, it is not suitable for the fatigue life assessment of bogie frames in the design stage.

3.3. PSD load spectrum

A PSD is composed of the load power spectrum density and frequency, which can represent dynamic loads in operational

Fig. 5. An example of three-dimensional stress spectrum.

7
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 6. An example of ASD [25].

scenarios and express most of the statistical information of random signals. It does not consider the probability distribution of time
signals. PSD is obtained by using a Fourier transformation [23] or other methods to convert the measured time domain signal in
operational scenarios to the frequency spectrum [24]. The PSD used for bench testing is usually expressed by an Acceleration
Spectrum Density (ASD), which is proposed by the International Electrotechnical Commission (IEC) [25]. An example of an ASD is
shown in Fig. 6. The process of generating an ASD can be found in references [25] and [26]. Note that a precondition for the use of a
PSD is that initial time signals should conform to the Gaussian distribution.

4. Simulated load spectrums

This section reviews simulated load spectrums of bogie frames generated from multibody dynamics and finite element (FE)
simulations; these simulated load spectrums can be used for bogie frame fatigue life assessments. Reviewed papers find that simulated
load spectrums include time domain load spectrum, frequency domain load spectrum and time-frequency domain load spectrum.

4.1. Time domain load spectrum

A time domain load spectrum is a stress spectrum, which is derived from the bogie frame’s load-time history as shown in Fig. 7.
After building a railway vehicle dynamic model, the time signal of track irregularity is inputted into the multi-body system (MBS)
software to obtain a load-time history (force-time history) at the bogie frames’ critical locations and force points (e.g., primary and
secondary springs). With the help of FE and rain flow counting techniques, the time domain load spectrum is obtained.
One critical step to generate the time domain load spectrum is to build a rigid-flexible coupled vehicle model which generally
consists of rigid bodies, flexible bodies and connection devices. The car-body, wheelsets and traction motors consist of rigid bodies.
The bogie frames consist of flexible bodies. Connection devices consist of springs, bumper stops and viscous dampers. Suspension
units can be linear or non-linear, which depends on the real vehicle data. Wheel-rail interaction is usually treated as non-linear. To
build a rigid-flexible coupled MBS, multi-body software should receive a bogie frame's model data from FE software. An example of
the data exchange between MBS and FE can be found in detail in [27] as shown in Fig. 8.
Luo et al. [17,28,29] established a four-axle railway vehicle dynamic model by using VAMPIRE software. Note that VAMPIRE is a
MBS simulation package which is specially designed for analysing the dynamic behaviour of railway vehicles [30]. Other multi-body
dynamic software packages have been developed for simulating mechanical systems (including railway vehicles). Stribersky et al.
[31] used SIMPACK to build a railway vehicle model. Beretta et al. [32] used ADAMS/RAIL [33] to simulate a railway vehicle system.
In some studies, only the front bogie frame is simulated as a flexible body to save computing time. This simplification would
inevitably cause some inaccuracy. Bogojevic et al. [34] and Nia [35] built a railway vehicle system using GENSYS [36] and validated
the vehicle dynamic model by using measured accelerations of the car body in the time domain. McClanachan et al. [37] used
constant friction dampers to build a wagon model by NUCARS.
Another important aspect of the time domain load spectrum is the transformation between load-time history and stress-time

Fig. 7. Process of time domain load spectrum.

8
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 8. Data exchange of FE-MBS.

history. Luo et al. [17] proposed the use of the FE technique to determine the bogie frames’ stress distribution. Firstly, unit loads are
applied to primary or secondary suspensions. Then, the bogie frame’s stress is calculated by FE software. Based on the correlation
coefficient between force and stress, the stress-time history in the critical areas is obtained. The single step of simulation for stress
calculation is described by [17,27,38]. Theoretical results of stress-time history are validated against field test data [17]. It confirms
that multi-body dynamic and FE techniques are reliable for obtaining dynamic loads acting on bogie frames.

4.2. Frequency domain load spectrum

A frequency domain load spectrum is usually presented as a stress spectrum, which is derived from the bogie frame’s PSD history
(PSD of force history) as shown in Fig. 9. MBS is inputted with the PSD signal of track irregularity to obtain the bogie frame’s PSD
history. Then, FE and Rayleigh techniques [41] (or Dirlik technique [39,40] and rain flow counting [10,46]) are used to obtain the
frequency domain load spectrum. It is worth mentioning that the PSD signal of track irregularity should obey the Gaussian stochastic
distribution in the frequency domain [42]. Details about how to obtain the bogie frame PSD history can be found in References
[43,44].
Stichel and Knothe [43] used MEDYNA to build MBS vehicle models. Younesian et al. [44] used LS-DYNA [45] to build MBS
vehicle models. Stichel and Knothe [43] verified that the PSD of stress has a good agreement with measured data. And they also
compared the PSD of force with the force spectrum obtained in the time domain. Results show that both of these have a good
agreement in the distributions of the suspensions’ forces. That means spectral analysis and covariance analysis seem to be useful for
studying bogie frame dynamic loads.

4.3. Time-frequency domain load spectrum

A time-frequency domain load spectrum is derived from the bogie frame’s time-frequency load history as shown in Fig. 10.MBS is

Fig. 9. Process of frequency domain load spectrum.

9
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 10. Process of time-frequency domain load spectrum.

inputted with the time signal of track irregularity to obtain the time domain load history. Then, a time-frequency transformation
technique (e.g., Fourier transformation) is used to obtain the frequency domain load spectrum. It is worthy of note that the bogie
frame’s load-time history should obey the Gaussian stochastic distribution [42] in the frequency domain. An example of obtaining the
time-frequency domain load history is described in [27].

5. Fatigue life assessment procedures for railway vehicle bogie frames

Bogie frame fatigue life assessment methods reviewed are categorised into the design load method, time domain method, fre-
quency domain method, time-frequency domain method, REPOS method and bench test method. This section reviews the fatigue life
assessment procedures of these methods.

5.1. Using standard-prescribed loads

The methods of using standard-prescribed design loads to assess bogie frame fatigue life are divided into pseudo-static (so-called
static) and cyclic load methods. This part reviews fatigue life assessment procedures of static and cyclic load methods.

5.1.1. Design (static) load method

The design (static) load method uses design loads (see part 2.1) prescribed by standards to assess bogie frame fatigue life as shown
in Fig. 11. Such design loads are first applied to bogie frame models. Then, numerical calculations are carried out by FE to obtain
stress distributions. Since the bogie frames’ stress distribution is a multiaxial stress state, it cannot be used for fatigue life assessment
directly. So multiaxial stress needs to be converted into a uniaxial stress state to obtain relevant fatigue parameters. Finally, the Haigh
diagram or Goodman diagram is used to assess bogie frame fatigue life. When mean stress and stress amplitude (or variable stress) are
inside of the diagram, bogie frame fatigue life can satisfy the design requirements. Fig. 12 shows examples of a bogie frame stress
distribution and a Haigh diagram for bogie frames. JIS 4207 [1] uses the Haigh diagram to assess the bogie frame fatigue life. UIC
515-4 [2], UIC 615-4 [3] and EN 13749 [4] use the Goodman diagram to assess the bogie frame fatigue life.
Existing methods of transforming multiaxial stress into uniaxial stress include the direct method[47], equivalent fatigue stress
method[48], equivalent mean stress method[49], projection method [50], amendatory Crossland method [51], tensor method [52]
and principal stress projection toward a spherical direction cosine group method [53]. Among all these methods, the projection
method introduced by ERRI [54] has become the most widely used method for railway vehicle fatigue life assessments [53].
For UIC 515-4 [2], UIC 615-4 [3] and EN13749 [4], the projection method can be described as follows. It assumes t load cases are
applied to the bogie frame. When in load case k, σmax occurs and its direction cosine is nk . In load case i , three principal stresses and
their direction cosines are defined as σ1i , σ2i , σ3i and n1i , n2i , n3i . The equivalent stress in load case i is calculated using Eq. (3). σmin is
the minimum value of σeq, i calculated for all load cases, except for load case k [53].
σeq,i = σ1i n1i ·nk + σ2i n2i ·nk + σ3i n3i ·nk (3)
Having obtained the maximum stress σmax and minimum stress σmin at each location, mean stress and stress amplitude can be

Fig. 11. Design (static) load method.

10
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 12. (a) Stress distribution of a bogie frame, (b) Haigh diagram [1].

calculated using Eq. (4) and Eq. (5) respectively.

σmax + σmin
σm =
2 (4)
σmax − σmin
∆σ =
2 (5)

For JIS 4207 [1], mean stress is the algebraic sum of stress calculated using each type of static loads. But, if there is a fluctuating
load, mean stress is the sum of the static load stress and half of the pulsating load stress. Variable stress (stress amplitude) is the
composite number of dynamic load stress, see Eq. (6).

σa = σ 21 + σ 22 + σ 23 + ⋯σ 2n (6)

Examples of using the design (static) load method to assess bogie frame fatigue life are as follows. Jeon et al. [55] and Oyan [56]
used JIS 4207 [1] to assess the bogie frame’s fatigue life. JIS 4208 [57] prescribes the static load test for assessing the bogie frame’s
fatigue life. Park et al. [52,44] and Kim [58] took advantage of in-service load conditions prescribed by UIC 615-4 [3] to assess bogie
frame fatigue life. Mancini and Cera [59] described in detail the design procedures, assessment methods, verification and manu-
facturing quality requirements of using EN 13749 [4] to assess the bogie frame fatigue life.
The design (static) load method neglects the effects of numbers of smaller cyclic loads and simplifies actual loads that occur in
operational scenarios. However, it is known that repeated cyclic loads may cause structural fatigue failure, even if stress levels are
substantially lower than static load stress [60]. To improve this situation, cyclic loads should be used for bogie frame fatigue life
assessments.

5.1.2. Design (cyclic) load method

The design (cyclic) load method uses cyclic loads (see part 2.2) to assess bogie frame fatigue life as shown in Fig. 13. Numerical
calculation is carried by FE to determine the stress distribution. Then, the S-N curve and Palmgren-Miner rule are used to assess bogie
frame fatigue life. Examples of using cyclic loads to assess bogie frame fatigue life are described in [64]. Details about fatigue tests
can be found in References [2–4,61–63].
The S-N curve (or Wöhler curves) characterises the relationship between stress level of material and fatigue life [65]. Jeon et al.

Fig. 13. Design (cyclic) load method.

11
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

[66] used the test to obtain the S-N of the GFRP composite bogie frame. There are some standards such as IIW [67], BS 7608 [68] and
EN 1993 [69] that also give S-N curves with several design classes for different welded joints. For one type of welded joint of one
material, the relationship between stress range and fatigue life can be expressed by Eq. (7).
S mN = C (7)
The Palmgren-Miner rule [70,71] is a widely accepted fatigue damage calculation method as shown in Eq. (8). When D = 1,

∑ Ni
failure is theoretically expected.

D=
n
i (8)
As the fatigue tests specified by UIC [2,3] and EN [4] have three load sequences, an equivalent stress range is usually used to
calculate the variable amplitude loading. The FKM guideline [72] describes how to use the Miner’s rule to calculate the equivalent
stress range for variable amplitude loading [64]. Combining Eq. (7) with Eq. (8), Eq. (9) can be derived to describe equivalent stress
range. Note that, in Eq. (9), the S-N curve is specified with two slopes; but the S-N curve which has one slope is also suitable for Eq.
(9).

1 ∑ Ni ·∆σi, s, d + ∆σR . L · ∑ Nj ·∆σ j, s, d ⎞ 1

m1− m2
σeq = ⎜⎛ ·
1
m1 m1

∑ Ni + ∑ Nj
m

⎟
⎝D ⎠ (9)
Only if the equivalent stress range is less than the fatigue limit, namely the results satisfy Eq. (10), would the bogie frame fatigue
life satisfy the design requirements [73–75].
σeq ≤ ∆σR, L (10)
For the design (cyclic) load method, it not only considers fatigue load amplitude and the number of load amplitude’s occurrence in
main load conditions, but also takes into account those cyclic loads which are larger but have relatively lower occurrence. The main
drawback of the design (cyclic) load method is that it cannot reproduce the vibration loads that bogie frames are subjected to in
operational scenarios. Therefore, load spectrums that accurately reproduce the operational scenarios are very important for bogie
frame fatigue life assessments [76].

5.2. Using measured or simulated load spectrums

The methods of using measured or simulated load spectrums to assess bogie frame fatigue life are divided into the time domain
method, frequency domain method, time-frequency domain method, REPOS method and bench test method. This part reviews fatigue
life assessment procedures of these methods.

5.2.1. Time domain method

In the time domain method, stress or strain time history is random vibration signals. The rain flow counting technique and
Palmgren-Miner rule are applied to assess bogie frame fatigue life. Fig. 14 shows the flow chart of the bogie frame fatigue life
assessment process by using the time domain technique. Note that the stress time history can not only be provided by the dynamic
stress test, but can also be given by MBS simulation. With the help of rain flow counting, the stress spectrum is obtained. And the
stress spectrum is a time domain load spectrum or dynamic stress test load spectrum (see parts 3.2 and 4.1).
Luo et al. [17] initially proposed the time domain method for bogie frame fatigue life assessments. What is more important is that
the theoretical calculation was verified by real-life of bogies in [17]. Beretta et al. [32] also used the time domain method to assess
the bogie frame’s fatigue life. The time domain method has a longer simulation time and higher accuracy. And sufficient samples and
regression analysis would make fatigue life assessment more confident [17], but this inevitably requires massive simulation time.

5.2.2. Frequency domain method

In the frequency domain method, spectral analysis by PSD intuitively describes the bogie frames’ stress distribution in service. The
PSD of stress is obtained by multi-body dynamics and FE techniques (see part 4.2). Then, Dirlik [39] (or Rayleigh [42]) technique,
rain flow counting and the Palmgren-Miner rule are applied to assess bogie frame fatigue life in the frequency domain as shown in
Fig. 15.
The frequency method for fatigue life assessment are reported in detail in [43,44,77–79]. References [43,44] use the Rayleigh

Fig. 14. Time domain method.

12
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 15. Frequency domain method.

technique [43,80] and Palmgren-Miner rule to assess the bogie frame fatigue life. The Rayleigh technique assumes that the load
signal is a narrow band, only considering positive peaks. Their valleys are matched with corresponding peaks of similar magnitude,
which may result in some stress cycles being over estimated. Using the Rayleigh technique to calculate fatigue damage may be over-
conservative [78]. If the rain flow counting technique is used for assessing fatigue life, results may be closer to actual. Younesian et al.
[44] compared the time domain method with the frequency domain method. Results show that the time domain method results in
higher safety factors in bogie frame designs.
The biggest advantage of the frequency domain method is that the simulations have fast computing speed. In similar accuracy
cases, the sampling rate of the time domain is about ten times higher than PSD in the frequency domain. Besides, the frequency
domain requires less storage space [77]. But limits of the frequency domain method are that the PSD must obey the Gaussian
stochastic distribution and the assumption that the railway vehicle system is linear. If deviation from the Gaussian distribution
occurs, related deviations of load spectra would happen. In these cases, the time domain method usually provides more accurate
results. Wolfsteiner and Breuer [81] proposed a method where non-Gaussian signals are decomposed in Gaussian portions. It regards
fatigue loads as the broad-banded signals. Then, Dirlik’s empirical solution for rain flow counting and the Palmgren-Miner rule are
used for fatigue life assessment in the frequency domain. The results verify that this method has a good agreement with the time
domain method for fatigue life assessment.

5.2.3. Time-frequency domain method

To reduce simulation time and have higher accuracy, Dietz et al. [27] proposed a time-frequency domain method for fatigue life
assessment as shown in Fig. 16. This method first utilises time domain or dynamic load tests to acquire the load-time history in the
time domain. Then, fatigue life assessment is performed in the frequency domain.
Details of the time-frequency domain method can be found in References [27,82]. The stress-time history is obtained by using
MBS and FE techniques (see part 4.1). Then, Rayleigh technique and Palmgren-Miner rule are used to assess bogie frame fatigue life.
The assessment for the bogie frame damage in the frequency domain can be accomplished using the nSoft fatigue software.
Compared with the frequency domain method, the time-frequency domain method is not restricted to a Gaussian distribution of
track irregularity. The time-frequency domain method retains the information of the time-domain statistical signal. When the
sampling rate is sufficient, the time-frequency domain method has higher accuracy than the frequency domain method. The fatigue
damage calculation of the time domain method is directly from the time signal. But the time-frequency domain method uses the
Rayleigh (or Dirlik) technique to express the PSD of stress corresponding to the time signal of the time domain method. Hence the
time domain method has higher accuracy. But the calculation time of the time-frequency domain method is shorter.

5.2.4. REPOS method

Use of the REPOS method for assessing the fatigue life of freight cars has been proposed by AAR [6]. REPOS is usually obtained by
field test and is rarely derived from the numerical technique. Based on the Palmgren-Miner rule, fatigue life cycles in REPOS can be

NT = 1/ ∑ αi/ Ni
calculated using Eq. (11).

(11)

when Smax > Se .

The number of cycles Ni to failure at given stress level i can be calculated using Eq. (12). Fatigue damage is defined as occurring

Ni = Ne /(Smax /Se ) m
1
(12)
Fatigue life is then determined using Eq. (13). If the railway vehicle is subjected to different load histories and different load
histories do not affect each other, Eq. (14) can be used.

Life (miles ) = NT /β ' (13)

Fig. 16. Time-frequency domain method.

13
 R. Xiu, et al.

Table 7
Comparison of fatigue life assessment methods.
Features Using prescribed loads Using measured or simulated load spectrums

Design (static) load method Design (cyclic) load method Time domain method Time-frequency domain method Frequency domain method REPOS method

Simulation time Short Short Long Normal Short Normal

Excitation Design (static) load Design (cyclic) load Load-time history (force) Load-time history(force) PSD of force Load-time history (force)
Limitation a. Unlimited life assumption Cannot reproduce real Sufficient random signal Time-frequency transformation a. Assumption of the track irregularity Force-stress

14
b. Cannot reproduce real operational scenarios samples obeys Gaussian distribution transformation
operational scenarios Assumption of railway vehicles is liner
Accuracy Conservative General Best Better Good Better
Theory Fatigue limit method Palmgren-Miner rule a. Rain flow counting a. Rayleigh technique a. Rayleigh technique a. Rain flow counting
Palmgren-Miner rule (or Dirlik and rain flow (or Dirlik and rain flow counting b. Palmgren-Miner rule
counting techniques) techniques)
Palmgren-Miner rule Palmgren-Miner rule
Engineering Failure Analysis 116 (2020) 104725
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

total Life (miles ) =

1/Life1 + 1/Life2 + ⋯
1
(14)

AAR [6] and Zarembski [8] illustrated the procedure of fatigue life assessment. The method for assessing fatigue life by REPOS
has been verified by laboratory testing. The numerical comparison of the fatigue life assessment method in operational scenarios is set
out in Table 7.

5.2.5. Other
The IEC [25] has proposed the use of a specific load spectrum (i.e., the ASD) to do the bench test to assess a bogie frame fatigue
life. The ASD load spectrum (see part 3.3) is based on the railway vehicle’s field testing. The advantages of this method are that it
repeats well, costs are lower and it takes less time. But the ASD is the real load spectrum for only one specific bogie frame, which may
result in fatigue assessment errors occurring.

6. Discussion

Standard-prescribed loads and measured or simulated load spectrums are usually used for assessing the fatigue life of railway
vehicle bogie frames. Standard-prescribed loads include design loads. For the design (static) load method, standards prescribe various
loads acting on the bogie frame. Dynamic load coefficients are used to represent those loads that occur frequently in normal operating
conditions. This may result in bogie frame fatigue life assessments that are not based on infinite life fatigue limit theory. The primary
disadvantage of the design (static) load method is that results are conservative. This is a reason why a design (static) load method is
usually used in the design stage or for remaining life assessment. The design (cyclic) load method not only considers load amplitude
and the number of load cycles, but also takes into account of those cyclic loads that have larger values with relatively lower oc-
currence. The cyclic load method improves the accuracy of bogie frame fatigue life assessments.
Measured or simulated load spectrums used for bogie frame fatigue life assessments fully consider the effects of track irregularity
on fatigue life. Using the measured load spectrum is the most direct and effective way, which has the highest accuracy. But its
disadvantage is that it is not suitable for the design stage and critical parts where strain gauges cannot be placed. Its higher cost and
longer test time result in that only a few bogie frames can be chosen for the field test. Use of a simulated load spectrum for fatigue life
assessment overcomes the drawbacks of the measured load spectrum. It is economic and time-saving. Many railway researchers use a
simulated load spectrum to assess bogie frame fatigue life in the design stage.
The time domain method uses a time domain load spectrum or dynamic stress test load spectrum to assess bogie frame fatigue life,
which has the longest simulation time and highest accuracy. The frequency domain method uses a frequency load spectrum to assess
fatigue life, which takes a shorter time. The main drawbacks of the frequency method are that it can only be used when the track
irregularity excitation obeys the Gaussian stochastic distribution and the vehicle systems are linear. The time-frequency domain
method uses a time-frequency load spectrum to assess fatigue life. It inherits the merit of time-saving from the frequency domain
method and has higher accuracy, similar to the time domain method. The REPOS method uses a REPOS load spectrum to assess
fatigue life, which can be performed manually without the use of a computer. It is always used in heavy haul freight locomotives.
Current research papers on assessing bogie frame fatigue life based on measured or simulated load spectrums do not introduce
fracture mechanics, resulting in the shortfall that crack propagation of bogie frames cannot be assessed. Further, the Palmgren-Miner
rule involved in the fatigue life assessment method does not take into account the effect of the load interactions, load sequences and
the damage from stresses below the fatigue limit.
As traction or braking forces applied to a powered bogie provide larger power for railway vehicles, their significant effect on bogie
frame fatigue life cannot be neglected, especially in high-speed and heavy-haul operational scenarios. Public literature indicates that
standards provide guidelines on how to apply loads and measured load spectrums should take into account traction and braking loads
under severe operational conditions. This means that the loads prescribed by standards (e.g., JIS 4207 [1] and EN 13749 [4]) cannot
reproduce operational scenarios; measured load spectrums that reproduce operational scenarios require longer simulation and
measured times, higher costs and are not suitable for the design stage. Simulated load spectrums only consider effects of track
irregularity, and other external loads (such as traction force, braking force and motor vibration) are neglected. Therefore, with
regards to using simulated load spectrums, there exists a gap in the knowledge of the application of traction or braking loads to bogie
frame fatigue life assessments.

7. Future work

Future work will investigate the previously identified research gap, i.e., current fatigue life assessment methods using simulated
load spectrums do not consider traction and braking forces. To fill this research gap, simulations have to consider traction forces,
braking forces and in-train forces. To do so, a number of approaches can be used. For example, the one shown in Fig. 17 uses
longitudinal train simulations to deliver in-train forces, traction forces and braking forces. Specifically, whole-trip train dynamics
simulations on a specific route will be carried out firstly. In-train forces, traction forces and braking forces will be saved and then used
in a MBS software (SIMPACK). Meanwhile, in the MBS model, track irregularity also has to be considered. To make the simulations
more realistic and accurate, FE bogie frame models can be incorporated into the MBS model. After the MBS simulations, a force-time
history at the bogie frames’ critical locations of secondary springs (or primary springs) and traction rod seats can be obtained. Having
obtained the force-time history, FE analysis and rain flow counting techniques can be used to obtain the REPOS of stress. Finally,

15
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Fig. 17. An example of fatigue life method considered traction loads.

relevant fatigue life calculation formulas (see part 5.2.4) can be used to assess bogie frame fatigue life.

8. Conclusion

A review of existing methods used for bogie frame fatigue life assessments is presented in this paper, particularly the fatigue life
assessment for locomotive bogie frames. These methods are using standard-prescribed loads and measured or simulated load spec-
trums to perform bogie frame fatigue life assessments. The methods of using standard-prescribed loads are divided into static and
cyclic load methods. The methods of using measured or simulated load spectrums are divided into the time domain method, fre-
quency domain method, time-frequency domain method, REPOS method and bench test method.
The design load methods (including static and cyclic load methods) are commonly used in the design stage or for remaining life
assessment. Standard-prescribed loads cannot truly reproduce the vibration loads in operational scenarios. Measured or simulated
load spectrums overcome the drawback of the standard-prescribed loads. Using measured or simulated load spectrums to assess bogie
frame fatigue life has a good agreement with its real fatigue life. Compromising the accuracy and analysis time, the time-frequency
domain and REPOS methods based on load spectrums are a good way to assess bogie frame fatigue life.

Declaration of Competing Interest

None.

Acknowledgment

The authors are grateful for the Centre for Railway Engineering, Central Queensland University, Australia.
This work is supported by China Scholarship Council and Natural Science Foundation of Changchun Normal University (Grant No.
2014008).

Appendix A. Notation

Notation Description

W Static load acting on one bogie frame

W1 Load caused by car-body mass acting on one bogie frame
W2 Payload acting on one bogie frame
W3 Load of the bogie frame’s mass and bogie frame components’ masses
W21 Vertical load caused by load-bearing masses’ vibration
W22 Vertical dynamic load caused by side beam mounted components masses’ vibration
W23 Vertical dynamic load caused by cross beam mounted components masses’ vibration
W24 Vertical dynamic load caused by end beam mounted components masses’ vibration
W25 Driving load acting on the driving seat in the vertical direction
W26 Braking load acting on the braking seat in the vertical direction
W27 Vertical damper load
W28 Anti-yaw load

16
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

WY1 Lateral dynamic load caused by load-bearing masses’ vibration and centrifugal force
WY2 Lateral bumper stop load
WY3 Lateral dynamic load caused by mounted components masses’ vibration
WY4 Lateral damper load
Wx1 Longitudinal load caused by load-bearing masses’ vibration
Wx2 Traction load
Wx3 Longitudinal load caused by mounted components masses’ vibration
Wx4 Braking load in the longitudinal direction
LP Load caused by mounted components’ masses
P Axle weight
f Friction coefficient between braking shoe and wheel tread and (or) between braking lining and braking disc
h Twist load
la Wheelbase
lp Primary spring lateral span
lr Wheel-to-rail contact point lateral distance
i Track twist value
Fz Vertical loads applied to each bogie
Fy Lateral loads applied to each bogie
Fx Longitudinal loads
Fzmax Maximum vertical loads applied to each bogie
Fymax Maximum lateral loads applied to each bogie
Fx1max Maximum longitudinal lozenging loads applied to each wheel
Ft1, Ft2 Twisting loads
mv Empty vehicle mass in running order
nb Number of bogies
C1 Passenger mass per seat (for passenger rolling stock), maximum payload (for freight rolling stock) and zero payload (for locomotives) in main
load cases
C2 Passenger mass per seat (for passenger rolling stock), maximum payload (for freight rolling stock) and zero payload (for locomotives) in
exceptional service
m+ Bogie mass
g Acceleration due to gravity
ne Number of axles per bogie
c Wheel loads of relevant bogie expressed as a %
fanti − roll Anti-roll torsion stiffness
FA Damper unloading force
Fz1, Fz2 Vertical loads applied to the side frames
α Rolling coefficient
β Bouncing coefficient
Fx1 Longitudinal lozenging loads applied to each wheel

Fzq1, −Fzq2
Fzs1, Fzs2 Static vertical loads

Fzd1, − Fzd2
Quasi-static vertical loads
Dynamic vertical loads
Fyq Quasi-static lateral load
Fyd Dynamic lateral load
σmax , σmin Maximum stress and minimum stress at each position
nk Cosine of σmax direction
n1i , n2i , n3i Direction cosine of three principal stresses in load case i
σ1i , σ2i , σ3i , Three principal stresses in load case i
σeqi Equivalent stress in load case i
σm Mean stress
∆σ Stress amplitude

σ1, σ2, σ3, ⋯σn

σa Variable stress
Stress corresponding to each dynamic load
S Stress amplitude
N Number of cycles corresponding to S
C Parameter related to materials, stress ratios, loading methods
m Slope of the S-N curve
D Structural damage
σi , ni i level of stress in the design stress spectrum and its number of cycles
l Total number of stress levels involved in damage calculation
σeq Structural equivalent stress
∆σi, s, d Design values of stress ranges above the knee point of the S-N curve
∆σj, s, d Design values of stress ranges below the knee point of the S-N curve
Nj Number of cycles at given stress range ∆σj, s, d
∆σR, L Design values of stress ranges (Resistance) at the knee point of the S-N curve
m1 Slope above knee point of the S-N curve
m2 Slope below knee point of the S-N curve
N0 Number of failure cycles corresponding to σ−1
NT Fatigue life cycles
αi Fraction of the total number of cycles for the given stress level i
Ni Number of cycles to failure at given stress level i
Se Fatigue limit stress

17
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

Ne Number of cycles of Se
Smax Maximum stress of the cycle
β' Total spectrum cycles per mile
Life1, Life2, Life3⋯ Lives at different load spectrum

References

[1] Japanese Industrial Standard: JIS E 4207, Truck frames for railway rolling stock-general rules for design, 2004.
[2] UIC Code 515-4. Passenger rolling stock-trailer bogies and running gear-bogie frame structure strength tests, 1993.
[3] UIC Code 615-4. Motive power units-bogies and running gear-bogie frame structure strength tests, 2003.
[4] EN 13749. Railway applications-Methods of specifying structural requirements of bogie frames. European Committee for Standardization, 2005.
[5] J.W. Cao, C.B. Xu, Application of engineering method for welding fatigue strength of locomotive bogie frame, Chinese J. Construct. Mach. 16 (1) (2018) 82–86
(Chinese).
[6] AAR specification M-1001: “Specifications for Design, Fabrication and Construction of Freight Cars”. Chapter 7: Fatigue Design of New Freight Cars, 1988.
[7] W. Schütz, A history of fatigue, Eng. Fract. Mech. 54 (2) (1996) 263–300.
[8] A.M. Zarembski, Freight car environment characterization for fatigue life analysis, Track/train Dyn. Des.: Adv. Tech. (1978) 205–221.
[9] R. Jones, J. Baker, Finding the Australian railway load spectrum design and assessment of light weight & durable railway structural components, in: E.E. Gdoutos
(Ed.), Fracture of Nano and Engineering Materials and Structures, Springer, Dordrecht, 2006.
[10] Y.L. Lee, M.E. Barkey, H.-L. Kang, Metal fatigue analysis handbook: Chapter 3: Rainflow Cycle Counting Techniques, Butterworth-Heinemann, 2012, pp. 89-114.
[11] H. Wu, P. Wu, F. Li, H. Shi, K. Xu, Fatigue analysis of the gearbox housing in high-speed trains under wheel polygonization using a multibody dynamics
algorithm, Eng. Fail. Anal. 100 (2019) 351–364.
[12] J.B. DeJonge, Fatigue load monitoring of tactical aircraft, NLR Report TR 69 0330, 1969.
[13] Z. Tang, P.B. Wu, G.F. Ding, J. Wang, Research on the vehicle bogie fatigue analysis and simulation platform, J. Syst. Simulat. 22 (8) (2010) 1813–1816
(Chinese).
[14] Q. Jia, Z.G. Cheng, A Fatigue life prediction of the bogie frame based on dynamic stress test, Railway Locomotive Car 30 (4) (2010) 65–67 (Chinese).
[15] Z. Ren, S. Sun, Q. Li, Z. Liu, Experimental studies of load characteristics of bogie frames for 350 km/h EMUs, Proc. Inst. Mech. Eng., Part F: J. Rail Rapid Transit
226 (2) (2011) 216–227.
[16] J. Li, J. Wang, X. Li, J. Yang, H. Wang, The experiment study for fatigue strength of bogie frame of Beijing subway vehicle under overload situation, Open Mech.
Eng. J. 9 (1) (2015) 260–265.
[17] R.K. Luo, B.L. Gabbitas, B.V. Brickle, Fatigue life evaluation of a railway vehicle bogie using an integrated dynamic simulation, Proc. Inst. Mech. Eng., Part F: J.
Rail Rapid Transit 208 (2) (1994) 123–132.
[18] L.X. Miao, S.G. Sun, P.M. Lv, Z.M. Liu, Q. Li, Test and research on stress spectrum for welded frame of speed increased passenger car bogies, Rolling Stock 36 (12)
(1998) 30–33 (Chinese).
[19] S.H. Baek, S.S. Cho, W.S. Joo, Fatigue life prediction based on the rain flow cycle counting method for the end beam of a freight car bogie, Int. J. Automot.
Technol. 9 (1) (2008) 95–101.
[20] P.B. Wu, S.T. Wen, J.B. Wang, Q.Q. Lan, Fatigue life prediction on the side frame of ZK2 bogie of the speed-increased freight car, China Railway Sci. 30 (1)
(2009) 91–95 (Chinese).
[21] J.S. Li, J.H. Wang, X. Li, J.W. Yang, H.T. Wang, The experiment study for fatigue strength of bogie frame of Beijing subway vehicle under overload situation,
Open Mech. Eng. J. 9 (2015) 260–265.
[22] W.H. Ma, S.H. Luo, R.R. Song, Load spectrum research of tumbler journal box node of high-speed motor car, Railway Locomotive Car 29 (4) (2009) 12–14
(Chinese).
[23] F. Hlawatsch, E. Auger, Time-Frequency Analysis, Wiley, Hoboken (NJ), 2005.
[24] E. Sejdic, I. Djurovic, J. Jiang, Time–frequency feature representation using energy concentration: An overview of recent advances, Digital Signal Process. 19 (1)
(2009) 153–183.
[25] IEC 61373. Railway applications- Rolling stock equipment-Shock and vibration tests, International Electrical Commission, 2004.
[26] A. Coulon, B. Nelain, N. Vincent, Interest of the transient accelerated testing approach on railway equipment qualification, Signal 2 (3) (2014) 4.
[27] S. Dietz, H. Netter, D. Sachau, Fatigue life predictions by coupling finite element and multibody systems calculations, in: Proceedings of DETC’97, ASME Design
Engineering Technical Conferences DETC/VIB, vol. 4229, 1997.
[28] R.K. Luo, B.L. Gabbitas, B.V. Brickle, Dynamic stress analysis of an open-shaped railway bogie frame, Eng. Fail. Anal. 3 (1) (1996) 53–64.
[29] R.K. Luo, Fatigue life evaluation of railway vehicle bogies using an integrated dynamic simulation, PhD thesis, South Bank University, London, 1993.
[30] M. Spiryagin, C. Cole, Y.Q. Sun, M. McClanachan, V. Spiryagin, T. McSweeney, Design and Simulation of Rail Vehicles, CRC Press, 2014.
[31] A. Stribersky, W. Rulka, H. Netter, A. Haigermoser, Modeling and simulation of advanced rail vehicles, IFAC Proc. Volumes 30 (8) (1997) 473–478.
[32] S. Beretta, F. Cheli, S. Melzi, M. Stanca, Dynamic simulation and estimation of stress spectra for railway components, in: 4TH ADAMS/RAIL Users’conference,
1999, p. 1-8.
[33] Manual of ADAMS/RAIL Multibody Dynamic Analysis Software Package, MSC.
[34] N. Bogojevic, P. Jonsson, S. Stichel, Iron ore transportation wagon with three-piece bogies–simulation model and validation, in: Proceedings of the 7th
International Scientific Conference Heavy Machinery (HM2011), Faculty of Mechanical Engineering, Kraljevo, 2011, p. 39-44.
[35] S.H. Nia, Dynamic modeling of freight wagons, Master’s degree thesis, BTH, Sweden, 2011.
[36] I. Persson, GENSYS in railway vehicle modeling, Manchester Benchmarks Rail Vehicle Simulation (2017) 91–100.
[37] M. McClanachan, C. Cole, D. Roach, B. Scown, An investigation of the effect of bogie and wagon pitch associated with longitudinal train dynamics, Veh. Syst.
Dyn. 33 (sup1) (1999) 374–385.
[38] S. Dietz, K. Knothe, Reduktion der Anzahl der Freiheitsgrade in Finite-Element-Substrukturen, Techn. Univ, ILR, 1997.
[39] T. Dirlik, Application of computers in fatigue analysis, PhD. thesis, University of Warwick, UK, 1985.
[40] R. Tovo, Cycle distribution and fatigue damage under broad-band random loading, Int. J. Fatigue 24 (11) (2002) 1137–1147.
[41] L.D. Lutes, M. Corazao, S.L.J. Hu, J. Zimmerman, Stochastic fatigue damage accumulation, J. Struct. Eng. 110 (11) (1984) 2585–2601.
[42] E.H. Vanmarcke, Properties of spectral moments with applications to random vibration, J. Eng. Mech. Div. 98 (2) (1972) 425–446.
[43] S. Stichel, K. Knothe, Fatigue life prediction for an S-train bogie, Veh. Syst. Dyn. 29 (S1) (1998) 390–403.
[44] D. Younesian, A. Solhmirzaei, A. Gachloo, Fatigue life estimation of MD36 and MD523 bogies based on damage accumulation and random fatigue theory, J.
Mech. Sci. Technol. 23 (8) (2009) 2149–2156.
[45] Y. Liu, ANSYS and LS-DYNA used for structural analysis, Int. J. Comput. Aided Eng. Technol. 1 (1) (2008) 31–44.
[46] N.W.M. Bishop, F. Sherratt, A theoretical solution for the estimation of “rainflow” ranges from power spectral density data, Fatigue Fract. Eng. Mater. Struct. 13
(4) (1990) 311–326.
[47] R. Li, Y.X. Zhao, Strength assessment for bogie frame based on UIC standard, Machinery 39 (10) (2012) 9–12 (Chinese).
[48] H.Y. Liu, P.P. Zhang, C.Y. Mi, Research of design method of rolling stock wheel strengths, J. China Railway Soc. 29 (1) (2007) 102–108 (Chinese).
[49] G. Sines, Behaviour of metals under complex static and alternating stresses, in: G. Sines, J.L. Waisman (Eds.), Metal Fatigue, New York, McGraw-Hill, 1959, pp.
145–169.

18
R. Xiu, et al. Engineering Failure Analysis 116 (2020) 104725

[50] B.H. Park, K.Y. Lee, Bogie frame design in consideration of fatigue strength and weight reduction, Proc. Inst. Mech. Eng., Part F: J. Rail Rapid Transit 220 (3)
(2006) 201–206.
[51] P.P. Zhang, J.C. Liu, B. Zhang, et al., Assessment method for fatigue strength of EMU wheel based on principal stress method and amendatory crossland method,
China Railway Sci. 35 (2014) 52–57 (Chinese).
[52] S.N. Xiao, C. Yang, G.W. Yang, A novel fatigue analysis method of welded frames, J. Southwest Jiaotong Univ. 48 (2) (2013) 199–204 (Chinese).
[53] B. Yang, B.Q. Ma, Y. Wu, et al., An improved projection method for determination of fatigue parameters of metal structures based on spherical direction cosine
group construction, Strength Mater. 48 (1) (2016) 127–134.
[54] ERRI B12/RP 17-1997. Programme of Tests to Be Carried Out on Wagons with Steel Underframe and Body Structure (Suitable for being Fitted with the
Automatic Buffing and Draw Coupler) and on Their Cast Steel Frame Bogies, 8th edition, European Rail Research Institute, Utrecht, 1997.
[55] K.W. Jeon, K.B. Shin, J.S. Kim, A study on fatigue life and strength of a GFRP composite bogie frame for urban subway trains, Proc. Eng. 10 (2011) 2405–2410.
[56] C. Oyan, Structural strength analysis of the bogie frame in Taipei rapid transit systems, Proc. Inst. Mech. Eng., Part F: J. Rail Rapid Transit 212 (3) (1998)
253–262.
[57] Japanese Industrial Standard: JIS E 4208, Test methods of static load for truck frames and truck bolsters of railway rolling stock, 1988.
[58] J.S. Kim, Fatigue assessment of tilting bogie frame for Korean tilting train: Analysis and static tests, Eng. Fail. Anal. 13 (8) (2006) 1326–1337.
[59] G. Mancini, A. Cera, Design of railway bogies in compliance with new EN 13749 European standard, in: Proceedings of WCRR (Vol. 2008), 2008.
[60] B.B. Muvdi, J.W. McNabb, Engineering Mechanics of Materials, Springer Science & Business Media, 1980, p. 137-147.
[61] B. Hwa Park, N. Po Kim, J. Seok Kim, K. Yong Lee, Optimum design of tilting bogie frame in consideration of fatigue strength and weight, Veh. Syst. Dyn. 44 (12)
(2006) 887–901.
[62] J.W. Han, J.D. Kim, S.Y. Song, Fatigue strength evaluation of a bogie frame for urban maglev train with fatigue test on full-scale test rig, Eng. Fail. Anal. 31
(2013) 412–420.
[63] Y. Zhang, P. Wu, Y. Song, Strength test and modal analysis for a standardized high-speed EMU motor bogie frame, in: 2015 4th International Conference on
Sensors. Measurement and Intelligent Materials. Atlantis Press, 2016.
[64] M. Kassner, Fatigue strength analysis of a welded railway vehicle structure by different methods, Int. J. Fatigue 34 (2012) 103–111.
[65] W. Schutz, A history of fatigue, Eng. Fract. Mech. 54 (2) (1996) 263–300.
[66] K.W. Jeon, K.B. Shin, J.S. Kim, A study on evaluation of fatigue strength of a GFRP composite bogie frame for urban subway vehicles, Adv. Compos. Mater 22 (4)
(2013) 213–225.
[67] A. Hobbacher, The IIW recommendations for fatigue design of welded joints and components, 2009.
[68] British standard 7608: code of practice for fatigue design and assessment of steel structures, 2014.
[69] British Standards Institution: BS EN 1993, Eurocode 3: Design of steel structures, 2006.
[70] J.E. Stallmeyer, W.H. Walker, Cumulative damage theories and application, Proc. Am. Soc. Civ. Eng. (1968).
[71] J.E. Stallmeyer, W.H. Walker, Cumulative damage theories and application, J. Struct. Div. 94 (12) (1968) 2739–2750.
[72] FKM-guideline- Analytical strength assessment of mechanical engineering components, VDMA Verlag, Frankfurt/M, 2002.
[73] B.F. Langer, Fatigue failure from stress cycles of varying amplitude, Trans. ASME J. appl. Mech 59 (1937) A160–A162.
[74] A. Palmgren, Die Lebensdauer van Kugellagern, VDI Zeitschrifft 68 (1924) 339-341.
[75] M.A. Miner, Cumulative damage in fatigue, Trans. ASME J. appl. Mech. 12 (1945) AI59-A164.
[76] F. Li, P. Wu, J. Zeng, C. Liu, et al., Vibration fatigue dynamic stress simulation under multi-load input condition: Application to metro lifeguard, Eng. Fail. Anal.
99 (2019) 141–152.
[77] N.W. Bishop, The use of frequency domain parameters to predict structural fatigue, PhD thesis, University of Warwick, 1988.
[78] N.W.M. Bishop, F. Sherratt, Fatigue life prediction from power spectral density data, Environ. Eng. 2 (1) (1989) 2–11.
[79] J.M. Tunna, Fatigue life prediction for Gaussian random loads at the design stage, Fatigue Fract. Eng. Mater. Struct. 9 (3) (1986) 169–184.
[80] C. Walck, Hand-book on statistical distributions for experimentalists (University of Stockholm Report No. SUF-PFY/96–01), 1996, p. 138-140.
[81] P. Wolfsteiner, W. Breuer, Fatigue assessment of vibrating rail vehicle bogie components under non-Gaussian random excitations using power spectral densities,
J. Sound Vib. 332 (22) (2013) 5867–5882.
[82] B.R. Miao, Simulation research of locomotive car body structure fatigue based on multibody dynamics and finite element method, PhD thesis, Southwest
Jiaotong University, China, 2007 (Chinese).

19