Vous êtes sur la page 1sur 14

Engineering Fracture Mechanics 98 (2013) 114

Contents lists available at SciVerse ScienceDirect

Engineering Fracture Mechanics


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

An investigation of the inuence of rail chill on crack growth


in a railway wheel due to braking loads
D. Peng a,, R. Jones a, T. Constable b
a
b

Department of Mechanical and Aerospace Engineering, Monash University, P.O. Box 31, Monash University, Victoria 3800, Australia
Asset Engineering, Operational Excellence, QR National, RC 1-11, 305 Edward Street, Brisbane 4000, Queensland, Australia

a r t i c l e

i n f o

Article history:
Received 12 March 2012
Received in revised form 25 November 2012
Accepted 1 December 2012

Keywords:
Tread braking
Rail chill effects
Stress intensity factor
Fatigue crack growth

a b s t r a c t
This paper provides a method for addressing thermal fatigue crack growth in railway
wheels, allowing for rail chill effects under block braking. A 3D non-linear nite element
model has been used to evaluate the thermal stress allowing for rail chill and for its inuence on the crack growth in a rail wheel. Both stop and drag braking are examined. The
results reveal that the effect of rail chill on crack growth is greater for drag braking than
it is for stop braking.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
It is widely believed that the most severe stresses in railroad car wheels are due to the non-uniform heating, which results
from tread braking. The current move towards higher train speeds and axle loads has created a demand for a deeper knowledge of the underlying science involved in wheel design. In Australia railway cars, for both passengers and freight, are quite
commonly braked by using blocks which contribute to the thermal load on the rail wheel. There are two kinds of braking:
stop braking and drag braking. Stop braking is used for stopping the train at scheduled stops or in an emergency. Drag braking is used to slow the trains over long distances in order to control speed. Brakes that are applied for long periods of time
result in prolonged heating of the tread and the rim. Thermal loading has the ability to initiate cracks, increase the crack
growth rate and reverse the residual stresses in the rim [1]. During tread braking process the heat generated by frictional
forces is distributed either as severe thermal gradients within the near surface or more gradually over the wheel rim and
plate, depending on the intensity of the heating source [2]. A number of in-service railway wheel failures associated with
this phenomena have been reported in [37]. The fatigue life of wheel treads has a strong bearing on the economy and safety
of rail transportation. Consequently an understanding of fatigue mechanisms and the ability to predict fatigue lives are of
interest to both manufacturers and operators. As a result there have been a number of studies into crack growth in railway
wheels [812].
Block braking of a wagon or a locomotive is performed by pressing the brake block(s) against the tread of the wheel,
which also is in rolling contact with the rail. The phenomenon of heat loss from a hot wheel through the contact area into
a cold rail is called rail chill. A simplied elastic analysis of an idealized braked rail-car wheel subjected to periodic brakeshoe thermal shock, rail chill and realistic tractive rail contact stresses has been used to demonstrate the important thermal
contributions to surface fatigue cracking using a critical plane fatigue initiation theory [13]. Some tests and experimental
Corresponding author.
E-mail address: daren.peng@monash.edu (D. Peng).
0013-7944/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engfracmech.2012.12.001

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

studies on the rail chill phenomenon have been reported in [14,15]. Simple analytical procedures and semi-analytical and
numerical methods for calculating the wheel tread temperatures for both non-perfect thermal contact conditions and perfect
thermal contact between wheel and rail are discussed in [1619]. An engineering tool, based on heat transfer principles, for
estimating the temperature variation of the wheel rim was reported in [20] which also presented a method for calculating
the rail chill effectiveness factor, i.e. the ratio of rail heat up and heat loss from wheel. In this work it was indicated that the
rail chill effectiveness is fairly small for stop braking and rather large for drag braking. Two-dimensional nite element models for the analysis of the temperature during railway tread braking (block(s) and wheel are coupled via a contact interface
that controls the heat generation and also the heat partitioning between the block and the wheel through thermal contact
resistances) have been provided and are presented in three parallel papers [2123]. This tool can be used for calculating temperatures during a full train route, including gradient braking and stop braking with intermediate cooling periods. There are
several ways to addresses the thermal behaviour of the blockwheelrail system and to determine the temperature elds in
the wheels and in the brake blocks. One approach is to use a prescribed frictional heat (introduced) at the blockwheel interface. The focus of the [21] was on the development of a numerical model used for analysing heat partitioning and temperatures in brake test rig experiments [22]. However, only wheel and block interaction was considered and the rail chill effect
was ignored. This approach was further developed in [23] to simulate the thermal interaction of wheel, blocks, and rail during braking and after cool down. It was stated that calibrated models can be used for studying changes in wheel behaviour
due to different block congurations and materials and that it can be used to study the difference in wheel behaviour between a brake test rig and operation in revenue trafc by accounting for the cooling inuence at the wheelrail contact.
The inuence of different wheel steel materials on the formation of the phase transformation zone of the wheel tread under
high speed and heavy thermal load conditions is discussed in [24]. This study involved a nite element analysis that allowed
for the characteristics of phase transformation.
Although a number of approaches have been used to study fatigue under braking loads, the authors could nd no references that addressed the effect of rail chill on crack growth in a rail wheel. As such the focus of this paper is to use a 3D nonlinear nite element model to evaluate the thermal stress with rail chill and to investigate its inuence on crack growth in a
rail wheel. This paper presents the results of a study into methods for estimating the thermal fatigue life of a S-shape plate
railway wheel under braking loading where the rail chill effect is accounted for in the analysis. This study consists of the
following areas of analysis. In the rst stage, a non-linear nite element analysis of the rail wheel allowing for braking
and the rail chill effect is performed. The second stage of the (sequential) analysis determines the stress intensity factor
for both small and large three dimensional (thermally induced) cracks using the method presented in [2528]. In the third
stage, the Generalised FrostDugdale approach [29,30] is used to model thermal fatigue crack growth. In order to consider
the inuence of rail chill effects on the crack growth, no mechanical loading or residual stresses (due to the manufacturing
process) are included in this analysis.

2. Thermal fatigue crack growth with rail chill effect model


2.1. Thermal load and rail chill
During the braking process, the friction generated by the brake-shoe on the moving tread produces heating. The heat generation at the braking shoewheel interface and heat transfer to the rail is shown schematically in Fig. 1. A break cycle includes breaking to full stop. As the wheel rotates points on the wheel tread, experience both heating and cooling. The
conditions at the two thermal contact interfaces which are associated with the wheelrail and brakewheel interfaces are
very complex. Many factors, such as roughness of the surfaces, oxides, lubricant, organic material and sand, may cause a
non-zero thermal contact (i.e. imperfect contact). In this analysis the interface thermal contact resistance has been ignored
and the widely used assumption [1013,2123] of perfect thermal contact between the two surfaces has been adopted.
At high Peclet number, the detailed distribution of the fast moving heat sources in the wheel tread is not important [31].
The most convenient and simple method is to assume that the mean heat ux is uniformly distributed around the wheel

Braking shoe

Wheel

Qshoe

Qwheel
Moving
Rail
Qrail
Fig. 1. Graphical representation of the rail chill effect.

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

circumference, thus determining the evolution of the mean temperature of wheel tread. It forms a major braking thermal
cycle. The assumption of a linear ramp loading pattern [11] is also common.
The braking heat transfer ux of a wheel can be estimated from:

qwheel t





Q wheel t 2Q wheel
t
Q0
t
wheel 1 

1
A1
tb
tb
A1
A1

where

Q wheel

PV 2
Q shoe
2gtb

Average instantaneous braking (friction) power (kW)

Q 0wheel 2Q wheel
Starting average braking power (kW)
In order to account for the dissipation of the thermal heat to the surroundings, a convectional boundary condition is applied to the surface of the wheel. As the wheel rotates the tread surface is subjected to periodic heating and cooling. A method to account for the cooling inuence from the rail is transformed to convection cooling to the area of wheelrail contact.
The cooling ux of rail chill effect is modelled using the equation:

qchill t 

Q rail
hrail T wheel t  T rail t
A2

Here Qrail is the heat into rail from wheel/rail contact surface and A2 is the contact area of wheel and rail, see Fig. 1. The
value of A2 can be obtained by using software such as GENSYS, and Vampire to model the wheel/rail contact conditions. In
this paper we used Hertzian contact theory, as a rst approximation, to evaluate contact area [32]. The value of hrail is the
coefcient of rail convection. The temperature of wheel and rail on the contact surface at time t are represented by Twheel(t)
and Trail(t) respectively. The rail convection heat transfer coefcient was taken from [21,33].
During braking longitudinal conduction has the effect of heating up a rail section ahead of the wheel. It thereby reduces
the rail chill effectiveness. It is found that longitudinal conduction can be negligible as the effect is minimal [20]. Consequently in this paper the effect of longitudinal heat conduction has been omitted from the rail chill analysis and Trail(t)
was assumed to be ambient temperature of the rail. The temperatures in the wheel resulting from the non-linear transient
analysis were then used to calculate the resulting thermal stress eld using non-linear nite element stress analysis.
2.2. Stress intensity factor and crack growth calculation method
This section presents the methodology used to model crack growth due to braking loads. The initiating aw was assumed
to a semi-elliptical surface aw (crack). As the thermal stresses generated by the braking loads are symmetrical and the
wheel has axi-symmetry, the growth of the surface crack at the plate/rim intersection was assumed to be normal to the tread
of the wheel. Furthermore, the depth and the surface length of the crack were allowed to evolve independently as the crack
grew.
Existing techniques for solving 3D crack problems for an arbitrary structure are either very time consuming, computationally intensive, computationally inefcient, or too restrictive. The problem is compounded when trying to establish the
fatigue life since the analysis must be repeated with a range of different crack shapes each requiring a separate and detailed
nite element mesh. To overcome this, a simple and accurate formulae for calculating the stress-intensity factors associated
with surface cracks at a stress concentrator, or notch, [2528] has been used. This methodology involves the use of a semianalytical approach that utilises the stress eld obtained from the uncracked nite element model. The advantage of
approaching the problem in this way is that it negates the need to explicitly model the crack in the nite element model.
In this way a crack of any size and (semi-elliptical) shape can be arbitrarily chosen and analysed using the original uncracked
nite element model. As the crack is not modelled explicitly a relatively coarse mesh can be used thereby minimising the
number of degrees of freedom, and improving analysis time. Solutions for the stress-intensity factors can then be relatively
quickly and easily obtained for a variety of cracks using the original uncracked nite element analysis.
Before proceeding to the results of the analysis, it is worthwhile outlining the method (used in the present paper) for
determining the thermal fatigue life of railway wheel under braking loads. The semi-analytical solution technique [25
28] was used to determine the stress intensity factors. The key component of this method involves determining the stress
intensity factors (SIFs) Ks using the equation:

h
i
a
KI K1
F e F s  F e ead q
I

Here K 1
I is the solution to the innite body problem obtained using the method given in [34], ad is a constant, q is the local
curvature, a is the depth of the surface semi-elliptical crack; Fe and Fs are the boundary correction factors taken from Newman and Raju [35], see [25] for more detail.

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

Fig. 2. Crack face geometry depicting non-symmetrical crack growth.

The rst step in the analysis is to model and analyse the region of interest. As outlined previously, the crack is not explicitly modelled. This uncracked nite element model is analysed. In this paper, we used NEi-NASTRAN [36] to determine the
stress eld in the uncracked structure. The stress eld on the crack face is then extracted and the Vijayakumar and Atluri [34]
method used to determine the innite body solution K 1
I : Eq. (3) is then used to compute the stress intensity factors around
the crack front.
Having determined a solution for the stress intensity factors for this particular aw shape, which is assumed to be semielliptical, the next step is to integrate the crack growth equation so as to determine the new crack shape. The Generalised
Frost and Dugdale [37] crack growth equation has been used to estimate crack growth. This equation1 presented in [29,30],
can be written in the form:


n
da=dN C  a1n=2 DK g K max 1g

where C, n and g are constants and Kmax is the maximum value of the stress intensity factor in a cycle. This formulation was
adopted since it has been shown to accurately represent crack growth in an extensive range of wheel steels [29] and accurately accounts for the growth of both small (sub mm) defects and large cracks under both constant amplitude and representative service loading [30,38].
Fig. 2 shows the crack face geometry used to describe the crack during propagation. The crack face is dened by two
parameters a, c the crack depth and half surface lengths respectively with two surface crack tips, labeled A and B,
at free surface. The non-uniform stress eld has the potential to create arbitrary stress intensity factors around the crack
front so that, in general, (K)a (K)A (K)B. Here, (K)a is stress intensity factor at the deepest point and the stress intensity
factors at the crack tips A and B are dened as (K)A and (K)B respectively. The values of t, WA,i and WB,i are used to determine
the boundary correction factors. Consequently, the growth of depth of the crack and each of surface tips must be considered
separately.
The crack growth increments Da and DcB and DcA are determined at each cycle. Since DcA and DcB can differ, the center of
the crack is allowed to move. A new center is determined through the assumption that the crack geometry will remain elliptical. New crack face coordinates are determined with the following equations:

ci1 ci

Dc A Dc B
2

ai1 ai Da
dshift;i1 dshift;i

5
6

Dc B  Dc A
2

These equations are evaluated at every step in the loading spectra to determine the next crack shape to be analysed.
3. Numerical examples and results analysis
3.1. Mesh, boundary conditions and thermal loads
This section presents the results of a study into methods for estimating fatigue crack growth in a 920 mm diameter rail Sshape plate rail wheel subjected to stop braking and drag braking. Particular attention is given to the inuence, on the crack
1

Derivations of Eq. (7) and its relationship to other crack growth equations are presented in the Appendix.

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

Fig. 3. The 3D mesh of the 920 mm rail wheel.

Thermal load band


(66 mm wide)

143 mm

The area of applied


rail chill flux loads
Fig. 4. Mesh local detail.

growth in the rail wheel, of the rail chill effect. Due to symmetry considerations only half of the wheel was modelled. The
resultant mesh, which was created using software program FEMAP [39], had 57980 nine-noded elements and 63096 nodes
(with 189288 degrees of freedom), see Figs. 3 and 4. To model the rail chill effect we used a ne mesh in the vicinity of the
wheelrail contact region. All of the surfaces of the wheel exposed to the atmosphere were assumed to experience free convective heat transfer. The initial model included both the wheel and the axel. This analysis revealed that the temperature at
the axel-wheel interface did not signicantly affected by allowing for the rail chill and was close to ambient. This conclusion
agrees with that reached in [40] and the approach recommended in the AAR standards [41]. As such, to reduce the analysis
time, subsequent analyses did not include the axel and assumed that the temperature at the wheelaxel interface was ambient and the nodes on wheelaxel interface were clamped in both radial and axial directions.
The rst problem analysed examined stop braking. In this case a 50.4 kW thermal load was applied to the tread region of
the wheel around its circumference. The train speed was taken to be 100 km/h and stop braking duration was 91 s. The thermal load band was 66 mm wide, to reect the use of a brake shoe. In service conditions, a small band of the wheel tread is
loaded in short fast repeated cycles. The heat source description was based on braking at a constant deceleration rate from an
initial speed to a full stop. The second case study involved drag braking with a rail chill effect. A 26 kW thermal load was
applied to the tread region of the wheel around its circumference. The thermal load was applied for a period of 1 h. The thermal load corresponds to approximately a 27 tonne axle load (i.e. 13.5 tonne per wheel). The thermal load was applied for the
duration, after which, the temperature variations throughout the wheel were calculated. These temperatures were subsequently used to calculate the resulting thermal stress eld in the wheel following thermal loading. The ambient temperature
was set at 25 C. As in [20] it was assumed that only 90% of heat generated goes into the railway wheel.
The area of applied rail chill ux loads was assumed to equal to the area of the wheel/rail contact patch. Hertz theory and
the VermeulenJohnson method [32] have been used to evaluate contact patch. The semi major axis (width of contact) and
semi minor axis of contact ellipse are 9.2 mm and 3.5 mm respectively.
3.2. Material properties
In this analysis, the material was assumed to be a high carbon steel wheel AAR grade B. For the initial verication study,
the material properties are assumed to be equivalent to Microalloyed AAR Class B wheel steel and are extracted from [42].
The room temperature yield strength of the material is set at 800 MPa, Youngs modulus at 206 GPa, Possions ratio at 0.286

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

and the density at 7870 kg/m3. The material thermal properties used are specic heat 490 J/kg C, coefcient of thermal
expansion 14  105, thermal conductivity 47.5 W/m C and free convection heat transfer coefcient 25  106 W/
C m2.
For the non-linear analysis, the effect of temperature on the stress strain characteristics is considered. The variations of
the yield strength and the elasticplastic property with temperature have been shown in Fig. 5. The coefcient of thermal
expansion, thermal conductivity and specic heat also varied with temperature and the functional form of these variations
were given in [42]. The software NE/NASTRAN [36] used for the non-linear nite element modelling allows those parameters
to be varied with temperature.

Normalized with the RT Value

Stress/Strain Characteristics viz. Temperature Range


1.2
Yield Strength
Elastic Modulus
Tangent Modulus

1.0
0.8
0.6
0.4
0.2
0.0
250

350

450

550

650

Temperature (C0)
Fig. 5. Variation in stress/strain characteristics over predicted temperature range.

Thermal Transient Analysis

Temperature (0C)

200
180
160
140
120
100
80
60
40
20
0

No Rail Chill Effect

10

20

30

40

50

Rail Chill Effect

60

70

80

90

100

Time (Seconds)
Fig. 6. Transient analysis result for the maximum tread region temperature of the wheel versus braking time for the stop braking application (average
power = 50.4 KW and duration = 91 s).

Fig. 7. Temperature prole in wheel under average 50.4 kW thermal loading at 37 s stop braking (with rail chill effect consideration).

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

3.3. Heat transfer transient analysis


Fig. 6 presents the temperature reached on the tread at different braking times for the stop braking case. The temperature
proles associated with the application of stop braking both with and without rail chill effects are shown in Fig. 7. The maximum temperature difference with and without the rail chill effect is approximately 11%. The temperature proles associ-

Temperature (0C)

Thermal Transient Analysis


650
600
550
500
450
400
350
300
250
200
150
100
50
0

No Rail Chill Effect

500

1000

1500

2000

Rail Chill Effect

2500

3000

3500

4000

Time (Seconds)
Fig. 8. Transient analysis result for the maximum tread region temperature of the wheel versus braking time for the drag braking application (average
power = 26 kW and duration = 3600 s).

Fig. 9. Temperature prole in wheel under average 26 kW thermal loading at 2700 s drag braking (with rail chill effect consideration).

Fig. 10. Non-linear thermal stress analysis results at wheel rim for the case of the stop braking with rail chill inuence.

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

ated with the drag braking case is shown in Fig. 8. In this instance there is a 9.8% maximum difference. Two meshes, one with
189288 (mesh 1) and 729534 (mesh 2) degrees of freedom respectively, were employed to check convergence. The maximum difference between these meshes was less than 1%.
Fig. 8 presents the maximum temperatures associated with drag braking at set locations in the wheel. After just 120 s of
heat input, the wheel rim reaches approximately 100 C. After 2700 s of thermal loading, it climbs to 632 C and 570 C for
the no rail chill and with rail chill effect respectively. The rim, tread and plate/rim intersection all increase in temperature.
Fig. 9 shows the temperature prole in the wheel under an average 26 kW thermal loading at 2700 s drag braking (with rail
chill effect). The difference in the maximum temperature between the no rail chill effect and with rail chill cases is approximate 9.8%.

Hoop Stress Amplitude (MPa)

A block of thermal cycles


400
Main Cycle

350
300
250
200
150
100
50

Sub-Cycles

0
1

51

101

151

201

251

301

351

401

Thermal Cycles
Fig. 11. A block cycles for the stop braking loading spectrum (included 1 main cycle and 407 sub-cycles).

Fig. 12. Non-linear thermal stress analysis results at wheel rim for the case of the drag braking with rail chill inuence.

Hoop Stress Amplitude (MPa)

A block of thermal cycles


350
300

Main Cycle

250
200
150
100
Sub-Cycles

50
0
100

5100

10100

15100

Cycles
Fig. 13. A block cycles for the drag braking loading spectrum (included 1 main cycle and 17600 sub-cycles).

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

3.4. Non-linear thermal stress analysis


In this analysis, 50 increments were used when performing the non-linear stress analysis. The thermal loading cycle that
the wheel is subject to through its life results in what is termed a stress reversal within the material. Over time the cyclic
loading causes the residual stress state of the wheel to change from a compressive stress state to tensile. This is due to the
small amounts of plastic deformation that occur during each thermal loading cycle. The material is unable to recover from
the plastic deformation, caused through thermal expansion when cooling occurs.
Fig. 10 shows the maximum hoop stress prole at a symmetric section of the wheel during the stop braking loading cycle
for the with rail chill effect. The maximum difference in the hoop stress distribution between the no rail chill effect analysis
and the with rail chill effect analysis was approximately 5%. Increasing the number of increments in the analysis to 60 resulted in essentially no difference in the results. In this convergence check, for the von-Mises stress and maximum principal
stress results, differences between mesh 1 (with 189288 degrees of freedom) and mesh 2 (with 729534 degrees of freedom)
were less than 2%. It was thus veried that the present mesh is sufciently accurate for estimating the thermal stress of the
wheel during tread braking.

Table 1
Crack growth estimating results for the wheel.
Braking

Rail chill

T (C)

Diff. (%)

rmax (MPa)

Diff. (%)

Stop

Yes
No

185
207

11

380
398

Drag

Yes
No

570
632

9.8

212
253

17

58.2 mm
a

a - Crack Depth
C - Crack Length

Fig. 14. The positions of the postulated cracks.

10

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

As cooling from the peak temperature down to room temperature occurs a tensile stress develops in the wheel tread. This
represents the main cycle seen during braking. However, in any one revolution, the temperature in the tread increases as
friction of block shoe and wheel occurs and reduces upon contact to the cool rail. Therefore within the 91 s braking process,
there are multiple sub-cycles that occur to the contact area of wheel/rail. To fully assess the crack propagation, both the main
cycle and all the sub-cycles may need to be considered. The maximum hoop stress level for the main cycle is approximately
380 MPa and for the sub-cycle the maximum stress level is approximately 25 MPa. The result of one block stop braking spectrum includes 1 main cycle and 407 sub-cycles, see Fig. 11.
Fig. 12 shows the hoop stresses in the wheel rim region of drag brake thermal loading for the with rail chill effect case. For
drag braking, there is 17% difference in hoop stress between the no rail chill and the with rail chill effect cases. Fig. 13
shows the block cycles for the drag braking loading spectrum (Included 1 main cycle and 17600 sub-cycles). The R-values
in both Figs. 11 and 13 are zero.
The differences in the maximum temperature and the hoop stress at rim due to rail chill are summarised in the Table 1 for
the two different load cases. Here, T and Smax are maximum temperature and hoop stress at rim respectively.
3.5. Fatigue life estimation
The material (micro-alloyed class B) properties, taken from [43], used to calculate crack growth were: n = 3.0, g = 1.0 and
C = 3.38  1012. The depth of the initiating defect was taken to be ai = 0.05 mm and the corresponding initial surface length
was ci = 0.2 mm. The denitions associated with these thermal induced initial surface defects [8,11] are shown in Fig. 14.
The resultant crack growth results are given in Figs. 15 and 16. The loading spectra used in this analysis are the normalized stress series listed in Figs. 11 and 13 for stop braking and drag braking respectively. For the wheel with Micro-alloyed
Class B material under stop braking, there is a difference of 18% in the computed fatigue life between the no rail chill and
the rail chill effect cases. For the drag braking case the difference is approximately 60%. This study revealed that the contribution of the sub-cycles to the total fatigue life was less than 5% for both stop and drag braking. It should be noted that the
crack initiation time is not included in this analysis. There could be a signicant number of extra cycles required to initiate
the 0.05 mm aw. A more detail crack initiation process analysis can be found in Ref. [13].
Basic aim of this paper is using 3D non-linear nite element model to evaluate the thermal stress with rail chill and investigating the inuence on the crack growth in the rail wheel. The loading due to the mechanical contact between wheel and

Crack Size vs. Number of Blocks


30
Without Rail Chill Effect, C

Crack Size (mm)

25

Without Rail Chill Effect, a


Rail Chill (No Sub-Cycle Effect), C

20

Rail Chill (No Sub-Cycle Effect), a

15

Rail Chill (Including Sub-Cycle Effect), C


Rail Chill (Including Sub-Cycle Effect), a

10
5
0
0

5000

10000

15000

20000

25000

Number of Blocks
Fig. 15. Crack propagation results for the wheel under stop braking loading spectrum at both direction of crack depth and crack length.

Crack Size vs. Number of Blocks


40
Without Rail Chill Effect, C

35

Crack Size (mm)

Without Rail Chill Effect, a

30

Rail Chill (No Sub-Cycle Effect), C

25

Rail Chill (No Sub-Cycle Effect), a


Rail Chill (Including Sub-Cycle Effect), C

20

Rail Chill (Including Sub-Cycle Effect), a

15
10
5
0
0

10000 20000 30000 40000 50000 60000 70000 80000 90000 100000

Number of Blocks
Fig. 16. Crack propagation results for the wheel under drag braking loading spectrum at both direction of crack depth and crack length.

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

11

rail was not included in this analysis. In our cases studied, it was shown that thermally induced cracking is essentially due to
braking loads with the mechanical loading playing only a small role and that rolling contact fatigue (RCF) appears to be dominantly due to the mechanical loads. The stress intensity factor at the maximum RCF crack size is much less than the fracture
toughness of the material. As such, in these instances, RCF does not increase the risk of failure directly. However, it may cause
shelling as the resultant crack enlarges and links with others, akes of material detach from the tread surface.
4. Conclusions
This paper has shown how heat transfer analysis and fracture mechanics tools can be used to compute the growth of complex three dimensional thermally induced fatigue cracks with and without allowance for rail chill effects. The analysis was
performed in three stages. In the rst stage, a 3D non-linear nite element model to evaluate the thermal stress with rail chill
and to investigate its inuence on crack growth in a rail wheel. Both stop braking and drag braking cases have been considered. The second stage of the sequential analysis is carried out to calculate the stress intensity factor of thermal cracks, by
using a semi-analytical solution technique that involves the use of an analytical solution combined with a numerical algorithm to assess fracture strength. In the third stage, a Generalised FrostDugdale approach has been used to modelling thermal fatigue crack growth.
In this analysis, the depth of the initiating defect was taken to be ai = 0.05 mm and the corresponding initial surface length
was ci = 0.2 mm. It is shown that for both cases the rail chill effect has a similar inuence (10%) on the maximum temperature in the wheel tread. It was also found that allowing for rail chill resulted in a difference of approximately 18% and 60% in
the fatigue life computed for the stop braking and drag braking cases respectively. The results of this analysis also reveal that
all sub-cycles can be omitted from the crack growth analysis as the damage due to these cycles is minimal.
It should be stressed that methodology used in this paper is ideal for use in railway wheel fatigue analysis as accurate
results are obtained without the need to explicitly model a crack, so it is only necessary to model the uncracked structure
(traditional analysis techniques require a ne numerical mesh to model the cracks, and the resultant structural fatigue problem becomes extremely time consuming and requires large computer resources).
Acknowledgement
This work was funded by the Commonwealth Research Centre for Railway Innovation through CRC project BR11.
Appendix A
It is now known [44] that fracture surfaces can be considered as an invasive fractal set. In this work Mandelbrot et al. [45]
wrote:
When a piece of metal is fractured either by tensile or impact loading the facture surface that is formed is rough and
irregular. Its shape is affected by the metals microstructure (such as grains, inclusions, and precipitates where characteristic length is large relative to the atomic scale), as well as by macrostructural inuences (such as the size, the shape of
the specimen, and the notch from which the fracture begins). However, repeated observation at various magnications
also reveal a variety of additional structures that fall between micro and macro and have not yet been described satisfactorily in a systematic manner. The experiments reported here reveal the existence of broad and clearly distinct zone
of intermediate scales in which the fracture is modelled very well by a fractal surface.
This concept, i.e. of a fracture surface as a fractal, has been further developed in [46,47] who used renormalisation techniques to develop a growth law for an invasive fractal, viz.

da=dN C 1 aDKn Ca/ DKn

A1

where n and / are constants. This law is formally identical to the classical Paris law except that the coefcient C1 depends on
the crack length a whereas in the Paris law it is assumed to be a material constant.
For small cracks DK can be expressed in the form:

p
DK bDr pa

A2

where b is the geometry correction factor which, since the crack is small, can be taken to be a constant. Consequently, for
small cracks, substituting Eq. (A2) into Eq. (A1) yields:

p
da=dN Cb pn a/n=2 Drn

A3

Let us now discuss a number of experimental observations that relate the crack growth rate to the crack length. The compendium of F/A-18 fatigue crack growth data by Molent et al. [48] examined more than 350 different cracks mainly in 7050T7451, but including some other 7000 series aluminium alloys, mill annealed Ti6Al4V titanium, and AF1410 steel that arose in a variety of full scale fatigue tests and associated coupon tests. In this study it was also found that in almost all cases
there was a near linear relationship, up to near failure, between the log of the crack length/depth and the number of load

12

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

blocks/ight hours and that this relationship held from a starting length of less than 100 lm to lengths in excess of 5 mms.
The papers [4953] have also shown that there is indeed a near linear relationship between the log of the crack size and life.
These observations are reinforced by the work of Polak and Zezulka [53] who studied the growth of physically small cracks,
which ranged in size from 20 lm to several mms in length, in austeniticferritic 2205 duplex stainless steel and a range of
other materials and concluded that over these length scales da/dN was proportional to the crack length and that the crack
growth history could be written as per Frost and Dugdale [52], in the form:

a ai e-N

A4

where - is a load and geometry dependent parameter and ai is the initial aw size. A near linear relationship between ln (a)
and N has also been reported, for physically small cracks, by a large number of researchers including Harkegard et al. [54],
Nisitani et al. [55], Kawagoishi et al. [56], Caton et al. [57], Murakamia and Miller [58], Polak and Zezulka [53], and Tomkins
[59]. This conclusion, i.e. a near linear relationship between ln (a) and N, was also one of the primary ndings reported by
Potter and Yee [60] in an early USAF/General Dynamics study into small crack growth undertaken as part of the F-16 research and development program. This nding held for the growth from small sub mm initial cracks up till near failure.
The (near) linear relationship between da/dN and ln (a), which appears to hold up to near failure, is a key feature of the lead
crack philosophy that is used to manage crack growth (up to failure) in RAAF combat aircraft [61]. The extensive data base
discussed in [62] which dealt with 3000 cracks in Chinese 840D freight wheels also which reported a near linear relationship (up to failure) between da/dTime and the crack length (a).
For Eq. (A3) to reect this near linear dependence of the crack growth rate on the crack length we require that, as a rst
approximation:

/ 1  n=2

A5

Substituting Eq. (A5) into Eq. (A1) we obtain:

da=dN C 1 aDKn Ca1n=2 DKn

A6

An alternative derivation which was obtained using notch mechanics is presented in [29]. This yielded the form of the
Generalised FrostDugdale crack growth equation used in the present paper, viz:

da=dN C  a1n=2 K gmax DK 1g n=1  K max =K c  da=dN0

A7

where C is a constant, the constant g controls the Kmax dependency of the crack growth rate da/dN, Kc is the apparent fracture toughness and the term (da/dN)0 reects both the nature of the initial defect as well as the apparent threshold for the
material. Here it should be noted that for crack growth from small initial defects in 7050-T7451 the threshold is very small
[63] so that the term (da/dN)0 can be neglected. The nding is echoed in the paper by Wanhill [64] who states that:
On the whole, it appears that fatigue crack growth thresholds are largely irrelevant for short/small ? long/large fatigue
crack growth, whatever the load history.

This formulation has been shown to accurately represent for the growth of small near micron cracks under both constant
amplitude [63] and operational loading [30,37]. Indeed this equation has been successfully used to predict crack growth under F/A-18, F-111 and a helicopter load spectra. It has also been shown to hold for a wide range of wheel steels [37] and a
range of other materials [43].
It is well know that there are numerous materials where the growth of long cracks exhibit a da/dN versus DK relationship
that is a function of the R ratio. In such cases it is common to dene an effective stress intensity factor range DKeff (=Kmax  Kop)
and a function U(R), which is a function of the R ratio, viz:

DK eff URDK DK  DK op

A8

DK op K op  DK min

A9

where

such that the various R ratio dependent da/dN versus DK (long crack) curves reduce to a single da/dN versus DKeff curve which
is (essentially) coincident with the high R ratio da/dN versus DK curve, see [61]. One such crack growth equation was presented in [65], viz:

da=dN gDK eff  DK eff ;th 2

A10

where g is a constant and DKeff,th is the threshold value of DKeff. In this formulation DKop is an exponential function of the
crack length so that for small cracks DKop becomes vanishingly small and da/dN becomes a function of (DK  DKth) as rst
proposed by Hartman and Schijve [66]. McEvily and Ishihara [67] subsequent revealed that for short cracks da/dN could be
approximated either as per Eq. (A10) or as a linear function of a Drn. This nding is supported by the review papers [39,68]
which reveal a large number of materials for which da/dN can be expressed as both as per Eq. (A7) [38] and as a function of

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

13

(DK  DKth)a [68], where a  2. Indeed, the ability of Eq. (A7) to reect the growth of cracks from small etch pits is shown in
[30,69]. An in-depth discussion of the relationship between the present crack growth equation, fractal based approaches and
Whlers and the Paris representations of fatigue is given in [70].
The ability of this formulation to predict the non-dimensional shape of the crack length versus cycles histories for a large
number of full scale fatigue tests, viz: the 1969 F-111 wing test, the Alohoa Airlines fuselage lap joint, the joint Switzerland
DassaultAustralia Mirage III full scale fatigue test, and an F/A-18 Centre Barrel test, as well as in coupon tests where the
specimens were subjected to complex operational load histories, viz: in the Boeing 757 and 767 materials characterization
tests, and the DSTO F/A-18 materials characterisation tests, is given in [43].
References
[1] Hackenberger DE, Lonsdale CP. An initial feasibility study to develop a wayside cracked railroad wheel detector. In: ASME/IEEE joint railway
conference, America; 1998.
[2] Cole I. The effect of localised cyclic heating on railway wheel steels an investigation into the effects of tread braking. In: 10th International wheelset
congress, Sydney; 1992.
[3] Fec CM, Sehitoglu H. Thermalmechanical damage in railroad wheels due to hot spotting. Wear 1985;102(12):3142.
[4] Peng D, Jones R, Constable T, Lingamanaik SN, Chen BK. The tool for assessing the damage tolerance of railway wheel under service conditions. Theor
Appl Fract Mech 2012;57(1):113.
[5] Gordon JE, Orringer O. Investigation of the effects of braking system congurations on thermal input to commuter car wheels, nal report of railroad
vehicle wheel performance (DOT-VNTSC-FRA/95-7). Volpe National Transportation Systems Centre, U.S. Department of Transportation; 1996.
[6] Ekberg A, Kabo E. Fatigue of railway wheels and rails under rolling contact and thermal loadingan overview. Wear 2005;258(78):1288300.
[7] Peng D, Jones R. The development of combination mechanical contact and thermal braking loads for railway wheel fatigue analysis. Theor Appl Fract
Mech 2012;60(1):104.
[8] Meizoso AM, Sevillano JG. Life prediction of thermally cracked railway wheels: growth estimation of cracks with arbitrary shape. Theor Appl Fract
Mech 1988;9(2):12339.
[9] Peng D, Jones R, Constable T. A study into crack growth in a railway wheel under thermal stop brake loading spectrum. Engng Fail Anal
2012;25(1):28090.
[10] Lunden R. Contact region fatigue of railway wheels under combined mechanical rolling pressure and thermal brake loading. Wear 1991;144(1):5770.
[11] Orringer O, Geay DE. Thermal cracking in railroad vehicle wheels subjected to high performance stop braking. Theor Appl Fract Mech 1995;23:5565.
[12] Donzella G, Scepi M, Trombini F. The effect of block braking on the residual stress state of a solid railway wheel. Proc Inst Mech Eng ProQuest Sci J
1998;212(2):14558.
[13] Moyar GJ, Stone DH. An analysis of the thermal contributions to railway wheel shelling. Wear 1991;144(2):11738.
[14] Stone DH, Carpenter GF. Wheel thermal damage limits. In: Proceedings of the 1994 ASME/IEEE joint (in conjunction with area 1994 annual technical
conference).
[15] Stone DH, Moyar GJ. Wheel shelling and spalling an interpretive review. Proc ASME Winter Annu Meet 1989.
[16] Moyar GJ, Carlson FG. Simple method for estimating railcar wheel tread temperature rise during braking. Rail Transport, ASME 1997;13:6574.
[17] Bejan A. Theory of rolling contact heat transfer. J Heat Transfer Trans ASME 1989;111(2):25763.
[18] Moyar GJ, Carpenter GF, Rajkumar BR. Heat transfer experiments with braked railcar wheels. ASME paper 86-WA/RT-1; 1986.
[19] Ertz M, Knothe K. A comparison of analytical and numerical methods for the calculation of temperatures in wheel/rail contact. Wear 2002;253(3
4):498508.
[20] Crowe KE, Raj PK. Analyses of rail chill effect. DOT/FRA/ORD-97-07, U.S. Department of Transportation; 1998.
[21] Vernersson T. Temperatures at railway tread braking. Part 1: modelling. Proc I MECH E Part F. J Rail Rapid Transit 2007;221(2):16782.
[22] Vernersson T. Temperatures at railway tread braking. Part 2: calibration and numerical examples. Proc I MECH E Part F. J Rail Rapid Transit
2007;221(4):42941.
[23] Vernersson T. Temperatures at railway tread braking. Part 3: wheel and block temperatures and the inuence of rail chill. Proc Inst Mech Eng Part F: J
Rail Rapid Transit 2007;221(4):44354.
[24] Hang SU, Tao P, Li L, Yang CF, Cui YH, JI HZ. Frictional heat-induced phase transformation on train wheel surface. J Iron Steel Res 2008;15(5):4955.
[25] Peng D, Jones R. An approach based on biological algorithm for three-dimensional shape optimisation with fracture strength constrains. Comput
Methods Appl Mech Eng 2008;197(4950):438398.
[26] Jones R, Peng D, Pitt S, Wallbrink C. Weight functions, CTOD, and related solutions for cracks at notches. Engng Fail Anal 2004;11(1):79114.
[27] Peng D, Wallbrink C, Jones R. Stress intensity factor solutions for nite body with quarter-elliptical aws emanating from a notch. Engng Fract Mech
2005;72:132943.
[28] Peng D, Wallbrink C, Jones R. An assessment of stress intensity factors for surface aws in a tubular member. Engng Fract Mech 2005;72(3):35771.
[29] Jones R, Pitt S, Peng D. The Generalised FrostDugdale approach to modelling fatigue crack growth. Engng Fail Anal 2008;15(8):113049.
[30] Jones R, Molent L, Pitt S. Crack growth of physically small cracks. Int J Fatigue 2007;29(911):165867.
[31] Tian X, Kennedy FE. Maximum and average ash temperatures in sliding contacts. ASME J Tribol 1994;116(1):16774.
[32] Johnson KL. Contact mechanics. Cambridge (UK): University Press; 1985.
[33] Mills AF. Heat transfer. Boston (USA): Richard D. Irwin, Inc.; 1992.
[34] Vijayakumar K, Atluri SN. Embedded elliptical crack, in an innite solid, subject to arbitrary crack-face tractions. J Appl Mech 1981;48(1):8896.
[35] Newman JC, Raju IS. Stress-intensity equations for cracks in three-dimensional nite bodies subjected to tension and bending loads. In: Atluri SN,
editor. Computational methods in the mechanics of fracture, vol. 2. Amsterdam (North-Holland): Elsevier Science Publishers; 1986.
[36] NE/NASTRAN Version 9.1. California (USA): Noran Engineering, Inc.; 2007.
[37] Jones R, Chen B, Pitt S. Similitude: cracking in steels. Theor Appl Fract Mech 2007;48(2):1618.
[38] Tiong UH, Jones R. Damage tolerance analysis of a helicopter component. Int J Fatigue 2009;31(6):104653.
[39] FEMAP Finite Element Modelling and Post Processing Version 9.3. Structural Dynamics Research Corporation, Pennsylvania; 2007.
[40] Ramanan L, Kumar RK, Sriraman RK. Thermo-mechanical nite element analysis of a rail wheel. Int J Mech Sci 1999;41(45):487505.
[41] AAR S-660-83. Procedure for the analytical evaluation of locomotive and freight car wheel designs. Manual of standards and recommended practices.
Mechanical Division Association of American Railroads, USA, G57; 1983.
[42] Mutton PJ, Choo P. Evaluation of low stress designs for 38 mm outstand railroad wheels. BHPR/ENG/R/91/009/CE07; 1991.
[43] Jones R, Pitt S, Constable T, Farahmand B. Observations on fatigue crack growth in a range of materials. Mater Des 2011;32(89):43628.
[44] Carpinteri Al. Scaling laws and renormalization groups for strength and toughness of disordered materials. Int J Solids Struct 1994;31(3):291302.
[45] Mandelbrot BB, Passoja DE, Paullay AJ. Fractal character of fracture surfaces of metals. Nature 1984;308:7212.
[46] Carpinteri A, Spagnoli A. A fractal analysis of size effect on fatigue crack growth. Int J Fatigue 2004;26:12533.
[47] Spagnoli A. Self-similarity and fractals in the Paris range of fatigue crack growth. Mech Mater 2005;37(5):51929.
[48] Molent L, Sun Q, Green AJ. The F/A-18 fatigue crack growth data, compendium, DSTO-TR-TR1667; February 2005.
[49] Barter S, Molent L, Goldsmith N, Jones R. An experimental evaluation of fatigue crack growth. Engng Fail Anal 2005;12(1):99128.

14

D. Peng et al. / Engineering Fracture Mechanics 98 (2013) 114

[50] Jones R, Barter S, Molent L, Pitt S. Crack growth at low Ks and the Dugdale law. J Chinese Inst Eng 2004;27(6):86975 [Special Issue in Honour of
Professor Sih GC].
[51] Molent L, Jones R, Barter S, Pitt S. Recent developments in fatigue crack growth assessment. Int J Fatigue 2006;28(12):175968.
[52] Frost NE, Dugdale DS. The propagation of fatigue cracks in sheet specimens. J Mech Phys Solids 1958;6(2):92110.
[53] Polak J, Zezulka P. Short crack growth and fatigue life in austeniticferritic duplex stainless steel. Fatigue Fract Engng Mater Struct
2005;28(10):92335.
[54] Harkegard G, Denk J, Stark K. Growth of naturally initiated fatigue cracks in ferritic gas turbine rotor steels. Int J Fatigue 2005;27(6):71526.
[55] Nisitani H, Goto M, Kawagoishi N. A small-crack growth law and its related phenomena. Engng Fract Mech 1992;41(4):499513.
[56] Kawagoishi N, Chen Q, Nisitani H. Signicance of the small crack growth law and its practical application. Metall Mater Trans A 2000;31(8):200513.
[57] Caton MJ, Jones JW, Boileau JM, Allison JE. The effect of solidication rate on the growth of small fatigue cracks in a cast 319-type aluminium alloy.
Metall Mater Trans A 1999;30(12):305568.
[58] Murakamia Y, Miller KJ. What is fatigue damage? A view point from the observation of low cycle fatigue process. Int J Fatigue 2005;27(8):9911005.
[59] Tomkins B. Fatigue crack propagationan analysis. Philos Mag 1968;18(155):104166.
[60] Potter JM, Yee BGW. Use of small crack data to bring about and quantify improvements to aircraft structural integrity. In: Proceedings AGARD
specialists meeting on behavior of short cracks in airframe structure, Toronto, Canada; 1982.
[61] Molent L, Barter SA, Wanhill RJH. The lead crack fatigue ling framework. Int J Fatigue 2011;33:32331.
[62] Xi NS, Liu JH, Liu XG, Liu FS, Wang JB. Fatigue failure and crack tolerance of 840D wheel web hole. In: Proceedings international heavy haul conference,
Specialist Technical Session (STS). Kiruna, Sweden; 2007. <http://www.ihha.net/conferences/past-conferences>.
[63] Jones R, Barter S, Chen F. Experimental studies into short crack growth. Engng Fail Anal 2011;18(7):171122.
[64] Wanhill RH. Characteristic stress intensity factor correlations of fatigue crack growth in high strength alloys: reviews and completion of NLR
investigations 19851990. NLR-TP-2009-256.
[65] McEvily AJ, Eier D, Macherauch E. An analysis of the growth of short fatigue cracks. Engng Fract Mech 1991;40:57184.
[66] Hartman A, Schijve J. The effects of environment and load frequency on the crack propagation law for macro fatigue crack growth in aluminum alloys.
Engng Fract Mech 1970;1(4):61531.
[67] McEvily AJ, Ishihara S. On the dependence of the rate of fatigue crack growth on the rn(2a) parameter. Int J Fatigue 2001;23:11520.
[68] Jones R, Molent L, Walker K. Fatigue crack growth in a diverse range of materials. Int J Fatigue 2012;40:4350.
[69] Burns JT, Larsen JM, Gangloff RP. Effect of initiation feature on microstructure-scale fatigue crack propagation in AlZnMgCu. Int J Fatigue
2012;42:10421.
[70] Marco PM, Carpinteri A. A unied interpretation of the power laws in fatigue and the analytical correlations between cyclic properties of engineering
materials. Int J Fatigue 2009;31(10):152431.

Vous aimerez peut-être aussi