Engineering Fracture Mechanics: Sciencedirect
Engineering Fracture Mechanics: Sciencedirect
Engineering Fracture Mechanics: Sciencedirect
A R T IC LE I N F O ABS TRA CT
Keywords: Railway axles with inside axle boxes have a great potential in modern urban railway system due
Fatigue crack growth rate to good dynamic performance on small-radius track. However, because of entirely different
Damage tolerance method loading nature from widely-used one, a challenging problem of the fatigue performance and
Inspection interval damage endurance arises in the case of defects from various external sources. In this paper, a
Fatigue damage accumulation
stepwise fatigue assessment framework composed of safe life as the first level and damage tol-
Modified Kitagawa-Takahashi diagram
erance as the second level was tentatively proposed for the wheelset with inside axle boxes. An
assumed load spectrum was employed to estimate the remaining life for a damaged axle made of
the EA4T steel. Calculated results show that the critical safety region has been transferred to the
axle center in contrast to classical axles, where a 1.0 mm-depth crack as the inspection limit has
less probability to propagate. Besides, the wheelset with inside axle boxes shows the smaller
stress value. The analysis from well-known Kitagawa-Takahashi diagram indicates a security risk
while only using nominal stress approach.
1. Introduction
Currently, an elaborate intercity railway system in China is being steadily established due to the depth industrialized process and
rapid economic development. This newly-designed track lines are also attempted to connect those existing moderate-speed or even
high-speed railway networks. Many interesting problems therefore arise besides the variable gauge as a fundamental issue. Among
the matched railway vehicles, those wheelsets assembled with inside axle boxes are recently proposed for more excellent dynamic
performance and curve passing as well as better riding comfort. However, one vital point of the assembled inside axle boxes should be
clearly remembered that it is the inside axle boxes placed on the axle that induce a radical difference in the longitudinal stress
distribution. In such case, the standardized strength design method and operation maintenance should be updated as quickly as
possible for the inside axle boxes based wheelsets.
It is well-known that the operation safety of the axle is always a principal concern, and the classical nominal stress based infinite
life design has validated a relatively conservative recommendation over the last few decades [1–3]. However, with the increased
lightweight and axle loads, progressive damages such as flying objects and corrosion pits happens externally in high-speed railway
components [4,5]. Such serious damage sites might lead to fatigue crack initiation and propagation under reversed cycle loads. To
ensure 25 years or more of the safe service, the axles therefore require to be strictly monitored and repaired by regular inspections
[4,6,7]. Recently, the fracture mechanics based damage tolerance complementary to nominal stress philosophy has been commonly
⁎
Corresponding author at: State Key Laboratory of Traction Power, Southwest Jiaotong University, 111, Section 1 North Second Ring Road, Chengdu 610031,
China.
E-mail address: [email protected] (S.C. Wu).
https://doi.org/10.1016/j.engfracmech.2018.04.046
Received 21 December 2017; Received in revised form 24 March 2018; Accepted 27 April 2018
Available online 04 May 2018
0013-7944/ © 2018 Elsevier Ltd. All rights reserved.
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Table 1
Main chemical compositions of medium strength EA4T steel grade (wt.%).
Cr Mn Mo Cu Ni C Si V
argued to evaluate the remaining life and resultant non-destructive testing (NDT) interval [8–10].
Unfortunately, compared with ample attentions on the axle with outside axle boxes [11,12], very few works up to now have been
devoted to investigating the fatigue behaviors and damaging influence of the axle with inside axle boxes. Despite of scarce experi-
ences, the stepwise state-of-art of assessing the axle in terms of nominal stress based safe-life design as the first level and fracture
mechanics based damage tolerance design as the second level can be fully developed for the newly-configured wheelsets. It is
therefore, for the first time, that current work focuses on the service capacity and durability of an axle assembled by inside axle boxes
on a mediate strength hollow axle.
In this paper, an open-published load spectrum as an example was adopted to validate the design methodology [8,13]. Small-scale
specimens based on low cycle fatigue, high cycle fatigue and fracture mechanics were adopted to extract the fatigue SN diagram and
fatigue crack growth data. The loading feature of the hollow axle with inside axle boxes was firstly simulated. Then, the total lifetime
from an evolved semi-elliptical crack (aspect ratio a/c = 0.8) located at the middle of the axle was numerically acquired by using a
newly-proposed crack growth model terms as LAPS [14], and the inspection interval was therefore suggested according to the
inspection limit.
The materials used for current European and Chinese high-speed railway axles include the low alloy standardized EA4T steel or
the quenched and tempered 25CrMo4 with medium strength grade [11,13–15]. The feasibility and capability of such steel grade into
modern urban rail transit system with internal axle boxes will be investigated in this work. The nominal chemical compositions in
percentage weight is listed in Table 1.
The material studied in Fig. 1 has a mixed microstructure of lamellar bainitic and lath martensite from the mid-thickness region of
hollow axles according to EN 13261, giving a micro-hardness of ∼206HV and relatively conservative properties. The average grain
size is 7.2 μm statistically estimated from electron backscattered diffraction (EBSD). Such fined grains generally indicates a good
deformation capacity and also a good fatigue short crack resistance [16,17]. For a standard monotonic tensile testing, the 0.2% offset
proof yield strength Rp0.2, an ultimate tensile strength Rm and a tensile fracture elongation can be calculated as 590 MPa, 704 MPa
and 18%, respectively.
Two types of hourglass specimens (see Fig. 2) were prepared to perform the experiments of both low cycle fatigue and high cycle
fatigue under axial tension-compression (TC) and rotating-bending (RB) loading. High cycle fatigued results based endurance curves
were then introduced to conduct a safe life assessment of railway hollow axles with inside axle boxes in terms of conventional
nominal stress method. However to ensure higher safety and reliability of service, the cyclic stress-strain responses of highly strained
region at the presence of a defect were measured to formulate a theoretical fatigue crack growth (FCG) rate model based on low cycle
fatigue behaviors[3,14,18]. Moreover, experimental long crack FCG rate (da/dN) and threshold stress intensity factor range (ΔKth)
under two stress ratios of R = −1 and 0 were also requisite with those Middle-crack Tension (MT) specimens. The selection of stress
Fig. 1. EBSD based grain orientation and texture of EA4T matrix from the mid-thickness axle.
177
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 2. Detailed shape and geometry of hourglass-shaped fatigue specimens (gauge length 30 mm and gauge diameter 6 mm) and plate-shaped
middle crack specimens (length 192 mm, width 48 mm and thickness 6 mm) from the mid-thickness of the hollow axle.
ratio is actually in view of carrying capacity of different locations of railway axles in terms of press-fitted effect and crack advance.
That is why all foregoing specimens should be carefully extracted from the axle close to the press-fits. Such vital region generally
represents the critical section of railway axles while designing according to European standards. With the influence of specimen
geometry, it is well identified that the MT geometry can achieve a similar FCG rate with relatively smaller scatter compared with that
from the Compact Tension (CT) specimens [19–21].
Both low cycle fatigue (LCF at strain rate 0.002/s) and axial high cycle fatigue (HCF with alternating sinusoidal loading mode)
tests were performed at a MTS 810 servo-hydraulic testing machine and an INSTRON 8872 high frequency fatigue machine, re-
spectively. The rotating bending HCF data under R = −1 were acquired from a PQ-6 rotating bending fatigue testing machine at
room temperature under sinusoidal loading frequency of 50 Hz. Such frequency value was selected for the frequency scope in axle
operation. It is assumed for high cycle fatigue testing that the specimen was manually terminated if the lifetime exceeded 107 cycles,
thus indicating a broken specimen.
As a necessary aspect and a primary parameter input to fracture mechanics based damage tolerance assessment, fatigue crack
propagation rate was recorded from above INSTRON 8872 fatigue testing machine in laboratory air under both R = −1 and 0. To
measure the fatigue crack length, the specimen surfaces were carefully polished to eliminate potential defects or short cracks. Upon
fatigue cycles, the crack propagates from the notch root, and the crack increment Δa and required total cycles N can be recorded. The
LCF properties based FCG curves were built without requiring fitted parameters from experimental FCG data and then compared with
those of standard NASGRO and classical Paris law.
in which Mrbm and σlim,rb are the bending moment and fatigue limit from full-scale axles with the seat diameter Dral and the bore
diameter dral, respectively. η is a generous security factor depending on all the uncertainties under entire operation conditions.
To acquire the fatigue strength limit of involved axles, HCF experiments are frequently preferred under operation environments.
The real load spectrum on interested sections is also acquired and then regularized to predict the fatigue life of railway axles with
inside axle boxes. However, those small-scale or laboratory specimens are always recommended in light of expense and complexity.
However, to tackle the geometry effect or so-called transferability between scaled specimens and expected full-scale axles under the
exactly same loading conditions, the nominal stress amplitude σa,lf of laboratory specimens in fatigue SN diagram has to be modified
by the relationship [24]
σa,fs = σa,lf αβεClot / Kt (2)
in which σa,fs is the predicted nominal stress amplitude of full-scaled axles. α, β, ε and Clot are empirically-determined factors for the
reliability coefficient, surface quality, geometry effect and loading type factors of small-scale specimens, respectively.
As an intrinsic characteristic of fatigue data, the scatter denoted via α should be carefully considered to correct the fatigue limit
178
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
from involved EA4T specimens. Thus, the fatigue diagram with the survival probability Pfal can be reduced to derive the remaining
life for equivalent full-scale axles by using the reliability coefficient α [12]
α = 1 + φ−1 (1−Pfal ) CVlfs (3)
in which φ and CVlfs are the standard normal cumulative distribution function and the variation coefficient for the fatigue limit at a
knee, respectively.
Currently, it is well-acknowledged that those stress levels far below the fatigue strength limit also play a very important role in
fatigue damage, inevitably leading to a finite life of the materials and structures. Such contribution is more evident especially for the
variable amplitude (VA) loading. The fatigue damage sum Dfad under a set of loaded cycles can be calculated according to the FKM
rules originally from Miner damage criteria. Thus the lifetime of railway axles with a measured load spectrum can be predicted once
Dfad approaches to a critical or allowable damage value as [12,22]
1 1
Dfad → Dcri = k
· ∑ ni ·σik + k ′ · ∑ ni ·σik ′
σa,fs Ni,f σi⩾ σa,fs σa,fs Ni,f σi⩽ σa,fs (4)
in which k and k′ = 2k − 1 are the slopes of fatigue SN diagram for the finite life and infinite life regime of the axle body, re-
spectively. Ni,f is the fatigue life at the i-level stress σi. In view of the effect of giga cycle fatigue range, the slope k′ = 2k − 2 of Wöhler
curves can be taken the fretting fatigue effect into account due to press fits [1,11,22,25].
(n ′+ 1) 1/(b + c )
1 ⎡ K ′εcy ΔK 2 ΔK 2 ⎞ ⎤
ΔN = · ·ln ⎛⎜
2 (σf′−σm ) ε′f (ΔK 2−ΔK th2 ) ⎝ ΔK th2 ⎠ ⎥
⎢ ⎟
⎣ ⎦ (6)
in which n′ and K′ are the cyclic strain hardening exponent and the cyclic strain hardening coefficient due to the LCF, respectively. b,
c, ε′f, σ′f and σm come from the Manson-Coffin equation. The subscript ‘cy’ represents the cyclic yield variables.
The material elements of the axle body are subjected to rotating bending loading typically at R = −1, which indicates that a
cyclic plastic zone ahead of the growing crack is gradually consolidated prior to unloading. Or it is said that the premature contact of
crack surfaces arises even at the tensile stage. On the other hand, the load interaction phenomenon arises due to complex dynamic
stress, which induces the local plasticity near the crack tip. In order to consider the plasticity-induced crack closure effect, a crack
opening function f can be used to modify both ΔK and ΔKth by [3,4,9]
1−f
ΔK eff = Kmax−K op = ΔK
1−R (7)
1−f
ΔKth,eff = ΔKth
1−R (8)
max(R,A0 + A1 R + A2 R2 + A3 R3) R ⩾ 0
f=⎧
⎨
⎩ A0 + A1 R −2 ⩽ R < 0 (9)
179
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 3. The wheelsets configuration, finite element meshes and loaded locations of railway hollow axles with inside axle boxes in case of a defect.
respectively. A0, A1, A2 and A3 are the coefficients of Newman function f. α is the constraint parameter related with stress status.
Precisely modeling an assembled wheelset with inside axle boxes is required for the stress field under real rotating bending loads
and press fits. By adopting the stress extrapolation method, the crack growth driving force ΔK at the deepest point of a growing crack
front (typically mode I component) can be evaluated after the axle 180° rotating. Finally, since the sudden fracture process usually
represents a very small portion of the axle lifetime, the critical crack depth or the stopping criterion for the damage tolerance is
defined to about 25 mm or risky 50% of net cross section in the simulation.
180
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 4. Calculated SIFs ahead of the crack tip with the increased crack depth inside the axle under different size ratios by using a stress extrapolation
method.
evolved crack depth a along the axial direction can be defined by h/a = 0.01, 0.05, 0.10, 0.15 and 0.20, respectively. Note here that
the stress extrapolation is selected rather than a complex J-integral approach, which is mainly due to its non-conservative nature
tightly depending on the integral paths. Such stress extrapolation method also permits the simple evaluation without prior as-
sumptions of the stress-train status.
It is obviously found from Fig. 4 that an acceptable or constant SIF prediction can be nearly acquired while the element size
around the crack tip is 1–10% of the evolved crack depth, thus giving a relative error below 10% or so that generally satisfies the
requirement for an engineering case. However, for a good balance between computational accuracy, efficiency and convergence, the
present work adopts the aspect ratio h/a = 0.05 to calculate the SIF values with a relative error 5% below, which is also consistent
with that widely used in similar damage tolerance assessment [9,10,13,19,25].
181
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 5. Definition of a selected 5-stage block loads used into fracture mechanics assessment corresponding to different service conditions [13].
spectra. However, it is actually impossible to elaborately conduct the durable and reliable assessment of the axles in a real en-
vironment by adopting such real loadings. Although the stress frequency distribution of the axle with inside box is rightly different
from the widely used, the purpose of this paper is actually to establish a general framework for evaluating the fatigue performance
and residual life of the axles with inside axle box. Here a five stage load spectrum (see Fig. 5) from open papers was adopted by a
reduced procedure, where the lowest load level was approximately equal to unit axle load [8,9,13].
It is clearly seen that different load levels enter into the load spectrum, which suggests that a probable crack closure might
frequently take place. Besides, the material elements inside central axle are susceptive to a fully reversed loading. Such synthetic
effect leads to an increased crack growth resistance partially from the overload. Therefore, a fatigue crack closure correction has to be
considered into the FCG model [4,6,9,14]. Moreover, the primary contribution of loading levels is that the extreme load can easily
buffer an initiation of a short crack even with a relatively small load frequency. Once a crack is initiated from a defect, the calculated
SIF range at the crack front increases rapidly. Other load levels start to be responsible for the subsequent crack growth. Such analysis
shows that the fatigue crack with several millimeters in depth actually controls the total lifetime of the axle. It is thus suggested that
the initial crack should be detected carefully.
For the residual lifetime assessment of the axles, it is necessary to determine the critical stress region under real load spectrum,
which directly determines the fatigue resistance. It is well-known for the alloy steel axles used in Europe and China that the stress
relief groove and geometrical transition are the most important position. Note here that the low carbon steel axles applied by
Shinkansen show the critical safety section exactly at the wheel seat because of small diameter ratio [2,6,25]. For those wheelsets
equipped by the outside axle boxes of intercity railway trains, the critical safety region of the axle (the transition between the wheel
and the gear seat) was clearly demonstrated in Fig. 6a under the peak load level F = 267 kN, as given in Fig. 5.
As a vital comparison, the longitudinal stress field within the wheelset newly-equipped with the inside axle boxes was also plotted
in Fig. 6b for validating the highly strained region of the axle. It is distinctly observed from Fig. 6b and c that no matter how the press-
fits were applied between the wheel, disc and gear and the axle, the peak stress region was always located at the central region of the
axle.
Moreover, the critical stress region of the newly-designed wheelset is fully different from the classical wheelset with outside axle
boxes beyond the two wheels. Note that the only difference between the two wheelset configurations was the location of two axle
boxes.
To further validate the interesting phenomenon, Fig. 7a firstly presents the longitudinal stress field of the wheel with the inside
axle box due to different loads. It can be clearly found that the central zone shows the largest stress value with the increased loads.
Fig. 7b then illustrates a distinct stress plot for two types of the wheelsets. Moreover, the press-fits cannot modify the stress value
especially at the middle of two axles. Besides, the peak stress level of the axle with inside axle box is considerably less than that of the
traditional axle. It is thus concluded that different stress distributions are generated even though a very slight change in the geometric
configuration of the axle box.
Previous results due to external axle boxes have shown that the critical stress location is caused by an integrated action from
rotating bending and press-fit loading [4,6,9], as shown in Fig. 6a. In contrast, the axle stress distribution with inside axle boxes was
only affected by rotating bending, and the press-fitted effect might provide a very tiny contribution, as clearly plotted in Fig. 6c. Such
interesting comparison in longitudinal stress fully argues that a dramatically higher level of nominal stress was generated at the
middle of axle body. Despite of little influence of press-fitted actions on the central axle, the stress relief groove between the gear and
wheel was still affected by the press-fits. The induced stress ratio of the material element (Fig. 6b) was found to be typical negative
182
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 6. Longitudinal stress plots under the same peak loading. (a) for the classical axle with press-fits, (b) and (c) for the assembled inside axle boxes
with and without press-fits.
unity, therefore presenting a dramatic closure action on fatigue crack growth under rotating bending mode. These findings also
suggest the site maintainer should pay more attentions on the central region especially when the external impacts easily induce a
defect larger than a = 0.5 mm in depth. Moreover, it is also indicated that the fatigue strength of different axle zones should be
specified when designing the axle assembled by inside axle boxes and selecting a suitable material for different speed grade of railway
trains.
Here the LCF properties were actually employed to formulate a FCG model LAPS as a comparison with standard NASGRO,
classical Paris law and experimental data rather not strain-life curves. By contrast, high cycle fatigue SN curves and resultant fatigue
strength limit from both axial TC and RB small-scale specimens were used to derive the fatigue performance of the axle in the
framework of safe life assessment.
183
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 7. Longitudinal stress at different positions of two axles. (a) inside axle box with press fit under load spectrum. (b) inside axle box and outside
axle box under the same peak load.
generally thought that a cyclic plasticity zone is locally consolidated before the crack penetrates the characteristic length. Such
globalized strain can well correlate with the localized phenomenon based FCG even under an elastic loading mode. Fig. 8 shows a
distinct strain-softening behavior in contrast with the monotonic tensile curve. The cyclic plastic parameters could be extracted as
n′ = 0.102, K′ = 806.30 MPa and σcy = 385.20 MPa. The fatigue ductility properties were obtained from Basquin equation as
b = −0.069, c = −0.641, σ′f = 811.1 MPa and ε′f = 0.658.
It is clearly found that the softening degree gradually decreases with the increasing of the cyclic strain amplitude. Moreover, the
material studied can be reasonably simplified with an ideal elastic-perfect plasticity. Thus the cyclic plasticity zone can be estimated
in terms of its dimension and morphology. The representative nonlinear elastic behaviors have been independently calculated by the
Fig. 8. Monotonic tensile and cyclic stress-strain curves of middle strength EA4T steel used for hollow axles with internal axle boxes of urban rail
transit train.
184
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 9. Axial TC fatigue SN curves with the failure probability of 2.5%, 50% and 97.5% for small-sized and derived full-scale EA4T specimens,
respectively.
well-known Hutchinson-Rice-Rosengren (HRR field) and Rice-Kujawski-Ellin (RKE field) [29,30]. For the HCF based long crack
growth, the elastic behavior dominates the singular stress field of crack tip. This cyclic strain-softening nature usually leads to a
smaller cyclic plastic zone ahead of a stable crack compared with those strain-hardened materials. Therefore, the material with strain-
softening would provide a faster FCG under the same loading. This paper adopts the cyclic constitutive parameters to formulate the
FCG model. However, it is still a complex problem to tentatively relate the LCF with fatigue crack growth and damage behaviors
[18,31,32].
Fig. 10. RB fatigue SN curves with the failure probability of 2.5%, 50% and 97.5% for small-sized and full-scale EA4T specimens.
185
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 11. Comparisons of predicted FCG curves via LAPS, NASGRO and classical Paris with those of experimental data of EA4T steel grade: (a) under
R = −1; (b) under R = 0.
Recent studies have shown that such simple calculations applying flexural beam principle are over conservative even under the
upper level of service load spectrum. This is partially attributed to the advent of intrinsic geometry discontinuity and extrinsic
defects. In light of the Cologne derailment in July 2008, the fracture mechanics estimations and resultant inspection intervals have
induced an intense discussion on life prediction and durability of operation axles. It is clearly identified that the determinant of a
verifiable residual life is preferably based on the framework of damage tolerance for a long crack. Fatigue crack growth rate has thus
become a standard step for the against fatigue design of railway axles. Due to routine maintenance and environment, the axle body
contains numerous crack-like defects, around which serious stress concentration occurs. Under the rotating bending load, such defect
might grow into a fatigue long crack once when the maximum SIF exceeds the threshold SIF. Fig. 11 shows the obtained FCG data as
well as analytical solutions from standard NASGRO and Paris, and newly-developed LAPS [14].
The comparisons of predicted curves clearly argue that both LAPS and NASGRO can well estimate the fatigue crack growth near
by the threshold region. In addition, the LAPS can acquire a slightly conservative prediction, which achieves a reliable remaining life
of a damaged hollow axle. By contrast, completely depending on artificially fitted parameters, standard NASGRO might consequently
accumulate a considerable life error, as illustrated by the upper and lower bounds in Fig. 11.
Such deviated FCG rates and resultant lifetime estimates clearly indicate the LAPS model without manually fitted parameters from
experimental FCG data has a distinct advantage over well-known NASGRO and Paris equations.
186
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 12. Different Miner rules into fatigue SN curves at 50% failure probability under measured load spectrum to the durability of the axles. Note
that the extended fatigue curve with the slope 2 k − 1 cannot really represent the lifetime variation larger than N = 107 cycles because gigy cycle
fatigue has indicated a stepwise tendency mostly due to interior cracking mode.
Figs. 9 and 10 give the probabilistic fatigue SN curves of full-scale powered axles with different failure probability of 2.5%, 50%
and 97.5%. Because of smaller scatter of testing data, the variation coefficient CVlfs = 0.048 to derive the reliability coefficient α for
the fatigue strength of full-scale axles at 2.5% failure probability [12,24]. The corrected fatigue limit σlim,rb of the axles under RB
loading was about 286 MPa at 50% failure probability, which was exactly equal to that of equivalent TC specimens. It is therefore
concluded that two types of small-scale specimens achieve a comparable result.
Based on infinite life design of powered axles made of steel grade EA4T with a security factor η = 1.66, the admissible fatigue
limits σasl of axle body were about 172 MPa. That derivation suggested that the maximum service stress should be lower than
172 MPa for those loaded motor axles. According to EN 13104, the calculated fatigue limit was larger than the prescribed 145 MPa.
Note here that the maximum permissible stress 145 MPa was obtained for the hollow powered axles made of EA4T steel grade by
introducing a security factor η = 1.66. Such result indicates the axle with inside axle boxes can reliably operate 30 years or have so-
called an infinite life under CA loading.
Table 2
Comparisons between calculated fatigue strength and service lifetimes from both small-scale specimens and derived full-scale axles via the MM rule
with 2.5% and 50% failure probability and critical damage sum Dcri = 0.3 and 0.5 by adopting an open load spectrum, respectively.
Specimens ND, cycles SD at failure probability Slope k Total cycles Lifetime, km
6 22
Small TC 5.71 × 10 335 MPa at 50% 22.4 1.75 × 10 4.74 × 1019
Small RB 2.65 × 106 369 MPa at 50% 11.8 2.33 × 1016 6.31 × 1013
Small TC 5.71 × 106 317 MPa at 2.5% 19.9 3.72 × 1019 1.01 × 1017
Small RB 2.65 × 106 357 MPa at 2.5% 13.4 7.26 × 1016 1.96 × 1014
Full TC 5.71 × 106 286 MPa at 50% 16.0 1.04 × 1023 2.80 × 1020
Full RB 2.65 × 106 286 MPa at 50% 8.60 7.17 × 1015 1.94 × 1013
Full TC 5.71 × 106 260 MPa at 2.5% 16.0 3.23 × 1021 8.83 × 1018
Full RB 2.65 × 106 260 MPa at 2.5% 8.60 9.19 × 1014 2.49 × 1012
187
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
From Table 2, it could be verified that the small-scale specimens surely achieved a higher fatigue strength and service lifetime
than those from derived full-scale axles. This reduced fatigue performance was effectively explained by the well-known geometry
effect of different material volume tested. On the other hand, according to the calculation results of derived full-scale axles, it could
be obviously found that the total lifetime was around 3.09 × 1012 km, which was exactly located into an infinite life design scope.
In other words, the railway axles with inside axle boxes would never fail during normal service subjected to the load spectrum
even though the peak load level is about 2.7 times of the axle load. These over conservative design and assessment results appear to
be unbelievable and strange in current engineering applications because some accidents have been recognized from the axle without
explicit fatigue damage.
This abnormal failure phenomenon is actually originated from the sub-surface cracking, which can be reasonably elucidated in
view of well-known very high cycle fatigue (VHCF) effect. Numerous investigations on the VHCF have indicated that some steel
materials can frequently fail at the load amplitude below the conventional fatigue limit or when the total lifetime exceed 107 loading
cycles [33–36], as demonstrated in Fig. 12. Actually the FKM Guidelines have to some degree considered the VHCF effect, therefore
showing adequate precision for approaching to the real life solution. However, a great deal of hard effort on improving the damage
accumulation has to be made for an economic design.
Fig. 13. Fatigue cracking growth driving force curves at the deepest point of a growing crack under real load spectrum for EA4Taxles with internal
axle boxes.
188
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
Fig. 14. Predicted life of the powered hollow axle with an initial crack depth 3.0 mm located at middle of the axle body in Fig. 6a.
and the critical crack depth were also plotted for a comparison with that of an undetectable defect in routine maintenance.
For an initial crack depth 1.0 mm as the inspection limit or a long crack less than the critical depth acri = 2.7 mm, a significant
crack arrest might take place under all load levels. On one hand, such locally damaged axle can be normally repaired by a suitable
polishing to prevent the crack growth. On the other hand, the crack growth simulations should be carried out once a crack-like defect
was detected for the residual life. It is therefore believed that once a crack generated at the middle of the axle is artificially overlooked
during normal maintenance, the prospective inspection interval of the damaged axle should be far less than the predicted lifetime.
From Fig. 14, the lifetime of the EA4T steel axle with inside axle boxes can be estimated from an initial crack depth 3 mm to the
half of wall thickness 25 mm, thus giving a total lifetime of 420 kkm or so. It should be noted that the fatigue crack would propagate
rapidly once it exceeds a = 5.0 mm, which corresponds to about 378 kkm. For an over conservative assessment, the remaining life
378 kkm should be adopted. By considering the safety factor of damage tolerance γ = 1.15 [12], the lifetime is corrected to 328kkm.
Thus a suitable inspection interval can be suggested from the corrected lifetime.
In one word, to assure the durability and reliability of the axles with inside axle boxes, the fracture mechanics based damage
tolerance analyses should be simultaneously performed as an important complementary to the nominal stress based infinite life
design. This result quantitatively suggests the technical attendants can establish an economic inspection interval rather than simply
replace the damaged axle by a new one.
in which ΔKth,lc and σa,th are the long FCG threshold and fatigue stress amplitude threshold of EA4T specimens with good surface
quality, respectively. a0,H is the El Haddad constant (=0.113 mm), and F is the geometry factor (=1.12) of cracked body [19,37].
On one hand, the nominal stress based fatigue limit assessment shows that the maximum permissible stress σa,fs = 260 MPa (see
Table 2) achieves the service lifetime 3.09 × 1012 km, generally meaning an infinite life scope. Based on Eq. (11), a maximum crack
depth a can be calculated about 0.57 mm, which is significantly less than the inspection limit 1.0 mm. This prediction means that an
initial crack at the depth a = 0.57 mm would never propagate based on Fig. 14. On the other hand, the critical crack depth
acri = 2.7 mm actually corresponds to a threshold stress σa,th = 64 MPa in terms of Eq. (11), which is dramatically less than the
maximum permissible stress σa,fs = 260 MPa. The result suggests that a long crack acri = 2.7 mm would rapidly propagate under
external stress 260 MPa. Besides, the calculated σa,th = 64 MPa is also lower than the estimated fatigue limit σasl = 172 MPa in terms
of a security factor η = 1.66 from Section 3.4.1. In other words, the smaller stress σasl = 172 MPa can still trigger the propagation
from acri = 2.7 mm. This analysis seems to seriously conflict with the infinite life assessment as validated in both Section 3.4.1 and
Table 2.
It is thus concluded that a potential security risk happens while using the safe strength and safe life assessment based on suitable
stress-life curves. In other words, the nominal stress approach cannot tackle the residual strength and life of a damaged axle. The
damage tolerance can provide a crucial complementary to the durability and reliability of the axles with inside axle boxes at the
presence of an external damage. The well-known KT diagram shows a huge potential to establish a good criterion of effective
inspection and assessment for normal operation of the train. Note that an elaborate study on the correlation between nominal stress
and fracture mechanics should be further made for the railway components especially when detecting a microstructural or physical
189
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
defect [19,37].
4. Conclusions
For the first time, by using the standard infinite life design method and advanced damage tolerance method, this paper tentatively
quantifies the durability and reliability of the axles with inside axle boxes subjected to an assumed load spectrum. Experimental data
of fatigue properties and fracture mechanics parameters were acquired with small-scale dog-bone specimens and middle-crack
tension plates. The fatigue crack growth rate was characterized by a newly-developed LAPS model from low cycle fatigue properties.
For the comparison and verification, standard NASGRO and Paris law were also introduced to fit experimental data. The remaining
life was evaluated and a possible inspection interval was suggested. Finally, the well-known Kitagawa-Takahashi diagram was in-
troduced to tentatively explain the results. Some points can be drawn as follows:
(1) Completely different from conventional hollow axles, the axles with inside axle boxes indicate that the critical safety region is
located at the middle of the axle primarily due to reversed loading mode.
(2) Compared with standard NASGRO, the newly-developed fatigue crack growth model termed as LAPS can achieve a valid and
intermediate remaining life result without requiring any artificially fitted parameters.
(3) Based on modified Miner rule to consider the damaging influence of small stress below the fatigue limit, the damage sum can be
suggested as Dcri=0.3 to assure the 30-year service life for the EA4T steel axle with inside axle boxes.
(4) For the non-destructive detection limit of current equipment, the initial crack depth a = 1.0 mm would not propagate even about
three times of axle load are applied on the axle based on damage tolerance approach.
(5) The calculated residual lifetime to a crack depth a = 5.0 mm is about 328kkm, and the inspection interval should be far less than
this predicted life for a damage axle once when an undetected crack arises.
(6) The Kitagawa-Takahashi diagram can effectively connect the safe life method as the first level with damage tolerance approach as
the second level, providing a staircase assessment framework of the axle with inside axle boxes.
Acknowledgements
The authors give very sincere thanks to the National Natural Science Foundation of China (11572267), the Sichuan Science and
Technology Program (2017JY0216), the Open Research Project of State Key Laboratory for Strength and Vibration of Mechanical
Structures (SV2016-KF-21), and the Open Research Project of State Key Laboratory of Traction Power (2018TPL_T03).
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.engfracmech.
2018.04.046.
References
[1] Smith RA, Hillmansen S. A brief historical overview of the fatigue of railway axles. Proc Inst Mech Eng Part F J Rail Rapid Transit 2004;218(4):267–77.
[2] Hirakawa K, Toyama K, Kubota M. The analysis and prevention of failure in railway axles. Int J Fatigue 1998;20(2):135–44.
[3] Wu SC, Zhang SQ, Xu ZW, Kang GZ, Cai LX. Cyclic plastic strain based damage tolerance for railway axles in China. Int J Fatigue 2016;93:64–70.
[4] Zerbst U, Beretta S, Köhler G, Lawton A, Vormwald M, Beier HT, et al. Safe life and damage tolerance aspects of railway axles – a review. Eng Fract Mech
2013;98(1):214–71.
[5] Goo JS, Kim JS, Shin KB. Evaluation of structural integrity after ballast-flying impact damage of a GFRP lightweight bogie frame for railway vehicles. J Mech Sci
Techol 2015;29(6):2349–56.
[6] Makino T, Kato T, Hirakawa K. Review of the fatigue damage tolerance of high-speed railway axles in Japan. Eng Fract Mech 2011;78:810–25.
[7] Kappes W, Hentschel D, Oelschlägel T. Potential improvements of the presently applied in-service inspection of wheelset axles. Int J Fatigue 2016;86:64–76.
[8] Pokorný P, Hutař P, Nahlík L. Residual fatigue lifetime estimation of railway axles for various loading spectra. Theor Appl Fract Mech 2016;82:25–32.
[9] Wu SC, Xu ZW, Liu YX, Kang GZ, Zhang ZX. On the residual life assessment of high-speed railway axles due to induction hardening. Int J Rail Transport 2018.
http://dx.doi.org/10.1080/23248378.2018.1427008.
[10] Traupe M, Jenne S, Lütkepohl K, Varfolomeev I. Experimental validation of inspection intervals for railway axles accompanying the engineering process. Int J
Fatigue 2016;86:44–51.
[11] Madia M, Beretta S, Zerbst U. An investigation on the influence of rotary bending and press fitting on stress intensity factors and fatigue crack growth in railway
axles. Eng Fract Mech 2008;75(8):1906–20.
[12] Beretta S, Regazzi D. Probabilistic fatigue assessment for railway axles and derivation of a simple format for damage calculations. Int J Fatigue 2016;86:13–23.
[13] Luke M, Varfolomeev I, Lütkepohl K, Esderts A. Fatigue crack growth in railway axles: assessment concept and validation tests. Eng Fract Mech 2011;78:714–30.
[14] Wu SC, Xu ZW, Yu C, Kafka OL, Liu WK. A physically short fatigue crack growth approach based on low cycle fatigue properties. Int J Fatigue 2017;103:185–95.
[15] Hassani-Gangaraj SM, Carboni M, Guagliano M. Finite element approach toward an advanced understanding of deep rolling induced residual stresses, and an
application to railway axles. Mater Des 2015;83:689–703.
[16] Deng GJ, Tu ST, Zhang XC, Wang QQ, Qin CH. Grain size effect on the small fatigue crack initiation and growth mechanisms of nickel-based superalloy GH4169.
Eng Fract Mech 2015;134:433–50.
[17] Wu SC, Xiao TQ, Withers PJ. The imaging of failure in structural materials by synchrotron radiation X-ray microtomography. Eng Fract Mech 2017;182:127–56.
[18] Madia M, Beretta S. An approximation for the cyclic state of stress ahead of cracks and its implications under fatigue crack growth. Eng Fract Mech
2011;78:573–84.
[19] Gänser HP, Maierhofer J, Tichy R, Zivkovic I, Pippan R, Luke M, et al. Damage tolerance of railway axles – the issue of transferability revisited. Int J Fatigue
2016;86:52–7.
[20] Vecchio RS, Crompton JS, Hertzberg RW. The influence of specimen geometry on near threshold fatigue crack growth. Fatigue Fract Eng Mater Struct
190
S.C. Wu et al. Engineering Fracture Mechanics 197 (2018) 176–191
1987;10:333–42.
[21] Hutař P, Seitl S, Knésl Z. Effect of constraint on fatigue crack propagation near threshold in medium carbon steel. Comput Mater Sci 2006;37:51–7.
[22] Traupe M, Landaberea A. EURAXLES-A global approach for design, production and maintenance of railway axles: WP2-development of numerical models for the
analysis of railway axles. Materialwiss Werkstofftech 2017;48(7):687–98.
[23] Cervello S. Fatigue properties of railway axles: new results of full-scale specimens from Euraxles project. Int J Fatigue 2016;86:2–12.
[24] Lee YL, Pan J, Hathaway R, Barkey M. Fatigue testing and analysis: theory and practice. Burlington: Elsevier Butterworth-Heinemann; 2005.
[25] Yamamoto M, Ishiduka H. Stress concentration of transition groove induced by a press-fitted part in railway axles. Int J Fatigue 2017;97:48–55.
[26] Zhu ML, Xuan FZ, Tu ST. Observation and modeling of physically short fatigue crack closure in terms of in-situ SEM fatigue test. Mater Sci Eng A
2014;618:86–95.
[27] Forman RG, Mettu SR. Behavior of surface and corner cracks subjected to tensile and bending loads in Ti-6Al-4V alloy. In: Ernst HA, Saxena A, McDowell DL,
editor. Fracture mechanics: 22nd symposium. Philadelphia: ASTM STP 1131; 1992. p. 519–46.
[28] Grubisic V, Fishcher G. Procedure for reliable durability validation of train axles. Materialwiss Werkstofftech 2006;37:973–82.
[29] Ellyin F. Fatigue damage, crack growth and life prediction. London: Chapman & Hall; 1997.
[30] Rice JR, Rosengren GF. Plain-strain deformation near a crack tip in a power-law hardening material. J Mech Phys Solids 1968;16:1–12.
[31] Zhu SP, Huang HZ, Ontiveros V, He L, Modarres M. Probabilistic low cycle fatigue life prediction using an energy-based damage parameter and accounting for
model uncertainty. Int J Damage Mech 2012;21(8):1128–53.
[32] Yang B, Ma BQ, Xiao SN, Zhao YX. A probabilistic model for describing short fatigue crack growth behavior of LZ50 steel. Strength Mater 2016;48(1):90–7.
[33] Lu LT, Zhang JW, Shiozawa K. Influence of inclusion size on S-N curve characteristics of high-strength steels in the giga-cycle fatigue regime. Fatigue Fract Eng
Mater Struct 2009;32(8):647–55.
[34] Zheng YC, Zhao ZH, Zhang Z, Zong WM, Dong C. Internal crack initiation characteristics and early growth behaviors for very-high-cycle fatigue of a titanium
alloy electron beam welded joints. Mater Sci Eng A 2017;706:311–8.
[35] Sun CQ, Xie JJ, Zhao AG, Lei ZQ, Hong YS. Acumulative damage model for fatigue life estimation of high-strength steels in high-cycle and very-high-cycle faigue
regimes. Fatigue Fract Eng Mater Struct 2012;35:638–47.
[36] Wang QY, Bathias C, Kawagoishi N, Chen Q. Effect of inclusion on subsurface crack initiation and gigacycle fatigue strength. Int J Fatigue 2002;24(12):1269–74.
[37] Maierhofer J, Pippan R, Gänser HP. Modified NASGRO equation for physically short cracks. Int J Fatigue 2014;59:200–7.
191