Effect of head wear on rail underhead radius stresses and fracture under high axle load conditions by Sagheer Abbas Ranjha B.Eng. (Hons), M. Eng. A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy Centre for Sustainable Infrastructure Faculty of Engineering and Industrial Sciences Swinburne University of Technology 2013
297
Embed
The effect of head wear on rail underhead radius stresses and ......Sagheer Abbas Ranjha B.Eng. (Hons), M. Eng. A thesis submitted in fulfilment of the requirements for the degree
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Effect of head wear on rail underhead radius stresses
and fracture under high axle load conditions
by
Sagheer Abbas Ranjha
B.Eng. (Hons), M. Eng.
A thesis submitted in fulfilment of the requirements for the degree of
Doctor of Philosophy
Centre for Sustainable Infrastructure
Faculty of Engineering and Industrial Sciences
Swinburne University of Technology
2013
Abstract
i
Abstract
Since the Hatfield incident in the UK in 2000, the possibility of catastrophic rail failure
as a result of the propagation of surface-initiated rolling contact fatigue (RCF) cracks
that go on to form transverse defects (TDs) is of on-going concern to the rail industry.
Generally, a rail can fail by either wear (loss of profile) or by RCF (surface cracks
growing by fatigue to cause rail break), and requires replacing. The heavy haul sector is
not immune to such concerns, although the combination of improved rail steels and
optimisation of the wheel-rail interface has reduced the extent to which rail wear
influences rail life. Field measurement results show a short duration tensile stress spike
at the underhead radius (UHR) of the rail as a heavy wheel passes over. This observed
stress is due to the lateral bending of the whole rail profile and the vertical and lateral
bending of the head-on-web. It is induced by the complex wheel-rail contact conditions
that are associated with, but are not limited to, a combination of wheel-rail contact
eccentricity from the rail centreline, lateral (transverse) forces and track support
conditions. In the presence of heavily worn rail, the magnitude of these stresses could
be even higher. Since the stress state is multi-axial with out-of-phase stress components
and varying principal stress directions, the cyclic wheel-rail contact, bending, residual
and thermally induced stresses interact with geometric features to produce a mixed
mode non-proportional stress history. This directly influences the location of fatigue
crack initiation, propagation and the severity of damage in rail with or without
macroscopic defects. Defect types that can be considered include, but are not limited to
head wear, reverse detail fracture, and long surface initiated rolling contact fatigue
cracks.
This study was conducted to parametrically evaluate the possibility of the failure risk
associated with rapid fracture behaviour of pre-existing RCF cracks and fatigue damage
at the rail underhead radius, and in particular, the qualitative assessment of the transition
Abstract
ii
to Mode I crack growth in the presence of a tension spike at the underhead radius. The
mechanical responses at the underhead radius were explored by modelling a single rail
on discrete elastic foundations using the finite element method (FEM). The wheel
contact load was modelled as a Hertzian contact pressure applied to an elliptical patch
on the rail head, assuming fully slipping conditions. The results revealed that the
longitudinal tensile stress at the rail underhead radius was highly dependent on several
operational parameters. The magnitude of tension spike increases and the region over
which tensile stresses occur moves closer to the top of the rail as a result of an increase
in the contact patch offset (CPO), the L/V ratio (ratio of lateral (L) to vertical (V) load),
head wear and the foundation stiffness. This stress is further increased by residual and
thermally induced stresses, and may result in a transition to Mode I crack growth for
pre-existing RCF cracks on the gauge corner of the rail that turn down to form
transverse defects (TD), as has been observed in practice [1-4]. These tensile stresses
may also result in fatigue crack initiation at the underhead radius [3-10].
The potential fatigue mechanism of crack initiation at the underhead radius was
implemented in a FORTRAN-code embedded in ABAQUS, as the damage parameter
defined by the Dang Van criterion. The Palmgren - Miner damage accumulation law
was used as an approach embedded in the FORTRAN-code to quantify damage
accumulation and cycles to failure. Hardness testing was conducted at the rail
underhead radius to estimate and quantify the fatigue properties. An examination was
conducted for different high strength rail steels, including the heat-treated low alloy,
eutectoid and hypereutectoid rail grades that are used under heavy haul conditions. The
fatigue behaviour of high strength rail steels was compared to predict the rail wear
limits and these results were correlated with field observations. As was reported in the
field [6-10], a reverse detail fracture is initiated in poorly lubricated, heavily worn
curved rail on stiff track carrying traffic under high axle load conditions.
A further study was conducted to investigate the unstable growth behaviour of single
and multiple pre-existing RCF cracks, especially the tendency for a rail break due to
rapid fracture under the high axle load conditions typical of those that exist in
Australian heavy haul operations. The occurrence of tension spikes as a result of the
Abstract
iii
lateral bending of the whole rail profile and localised vertical and lateral bending of the
head-on-web is exacerbated with increasing rail head wear and the propensity of rapid
fracture associated behaviour correlates with the extent of rail head wear. This
behaviour was examined using the extended finite element method (X-FEM), to
parametrically study the unstable growth behaviour of RCF cracks. The extended finite
element method results revealed that existing RCF cracks, when subjected to high
tensile stresses at the underhead radius, could contribute to the development of a rapid
(unstable) fracture. The results of this thesis can be used to examine the influence of
wheel-rail interaction behaviour and rail head wear on the possibility of a catastrophic
rail failure developing from RCF damage.
This page is left intentionally blank
Declaration
v
Declaration
I declare that no portion of the work referred to in the thesis has been submitted in
support of an application for another degree or qualification of this or any other
university or other institution of learning. I also declare that to the best of my
knowledge it contains no material previously published or written by another person
Appendix A. Equations for rail bending stresses ....................................... 221
Appendix B. Rail underhead radius (UHR) stresses data .......................... 227
Appendix C. FORTRAN-code based on the Dang Van fatigue criterion .. 261
List of Figures
xix
List of Figures
Chapter 1
Figure 1.1 Appearance of RCF damage on high rail [2] 2
Figure 1.2 Development of transverse defect [1] 3
Figure 1.3 Transverse defect (TD) developed from RCF damage [1, 2, 16, 17]
3
Figure 1.4 āWhole Life Rail Modelā āWLRMā (a): Crack propagation rate v/s length (adopted from [1, 20, 22-24, 30, 31]); (b): Sequence of events in the development of RCF [22, 25, 26]
5
Figure 1.5 Assumed longitudinal bending response of rail in āWhole Life Rail Modelā [27]
7
Figure 1.6 Measured longitudinal stress response under in service loading; example shown for gauge side underhead radius of high rail in 609 m radius curve [1, 32, 33]
7
Figure 1.7 Representation of localised head-on-web bending and lateral bending of the whole rail profile as a beam model on a continuous elastic foundation with eccentric vertical and outward lateral loading (adopted from Orringer et al [46, 47], Jeong et al [7], and Salehi [49])
9
Figure 1.8 RDF development [16] 11
Figure 1.9 RDF (Courtesy of TSBC [3, 9]) 11
Figure 1.10 RDF development [16] 11
Figure 1.11 RDF (Sperry rail services [6-8]) 11 Figure 1.12 RDF development in aluminothermic welded rail [50] 11
Figure 2.5 Hertzian pressure distribution at the contact patch [the picture to the left is from , whereas the picture to the left is from
24
Figure 2.6 (a) Vehicle on a curve; (b) Rolling radius difference [64] 25
Figure 2.7 Lateral shift of wheel set leads to contact patch offset from the center of railhead cross-section [66]
26
Figure 2.8 Representation of surface traction on elastic half space [58] 28
Figure 2.9 (a) Stress distribution versus depth caused by Hertz pressure acting on a circular area of radius āaā [58] (b) subsurface stresses; (c) contours of principal shear stress [58]
29
Figure 2.10 Material response to repeated loading [73] 30
Figure 2.11 Shakedown map for repeated sliding of a rigid cylinder over an elastic-plastic half space [74, 75, 77]
31
Figure 2.12 Shakedown diagram and RCF predictions [79] (high rail on a moderate curved track with axle load of 30 tonne)
33
Figure 2.13 Photomicrograph of an etched sectioned test sample from twin disc test [88]
35
Figure 2.14 Schematic of the mechanism of RCF crack formation and branching [91]
36
Figure 2.15 Broken rail showing shell origin of detail fracture [Courtesy of Transportation Technology Centre, Inc. (TTCI)] [8, 10]
37
Figure 2.16 Predicted head stresses in 60 kg/m and 68 kg/m rails [36] 40
Figure 2.17 Effect of contact patch offset (CPO) on longitudinal stresses at underhead radius (UHR) for 68Kg/m rail at nominal axle load of 30 and 35 tonnes, dynamic factor of 1.5 and L/V ratio of 0.2 [37]
41
Figure 2.18 Predicted rail head stresses for different head designs [36] 41
Figure 2.19 Fatigue prediction loading paths at different points of the rail surface [52, 53, 99]
46
Figure 2.20 Representation of critical plane with shear stress at a material point in a shear plane defined by normal vector ānā, coordinate system xyz (adopted from Ekberg et al [62])
47
Figure 2.21 Modes of crack growth (a) Mode I, opening (b) Mode II, shearing (c) Mode III, tearing [20]
48
Figure 2.22 Three phases of the life of a RCF crack initiated at the surface [95, 114]
49
Figure 2.23 Fluid assisted mechanisms of crack growth: (a) shear mode crack growth, accelerated by fluid reduction of friction between the crack faces, (b) hydraulic transmission of contact pressure, (c) entrapment and pressurization of fluid inside the crack and (d) squeeze film pressure generation [117, 118].
50
List of Figures
xxi
Figure 2.24 Effect of wear on crack growth [22, 23] 54
Figure 2.25 Life cycle of a crack in a rail [125] 55
Figure 2.26 Strategies for rail life management [22, 23, 25] 56
Figure 2.27 Hardness distribution along vertical transverse at rail centerline [2]
59
Figure 2.28 Relationship between yield strength and hardness in eutectoid and hypereutectoid heat treated rail steel [2]
60
Figure 2.29 Relationship between tensile strength and fatigue limit [130] 60
Chapter 3
Figure 3.1 68 kg/m rail [80] 64
Figure 3.2 Flow diagram for FE model solution technique development based on rail geometry, material properties, loading and boundary conditions
65
Figure 3.3 The model description: (a) Top view of single rail supported with discrete elastic foundations; (b) Cross-section view of the rail, the measurement location and the elastic foundations
67
Figure 3.4 Variations of contact patch sizes and shapes 70
Figure 3.5 Representation of a contact patch offset (CPO) from the center of rail head cross-section for modelling eccentric loading
71
Figure 3.6 Representation of inward and outward lateral shear tractions by leading and trailing wheelsets respectively during curving on high rail of a 600 m radius curve (adopted from Xiao et al [134])
72
Figure 3.7 The rail-wheel contact loads in detail (a) wheel loads [the picture to the left is the original work]; (b) Hertzian pressure distribution at the contact patch [picture to the right is adopted from Ekberg et al [62]]
73
Figure 3.8 Mesh development with dense mesh at the contact patch 74
Figure 3.9 The strain gauges at the measured points [33] 75
Figure 3.10 Measured longitudinal stress response under in service loading; example shown for high rail in 609 m radius curve [1, 32]
76
Figure 3.11 Stress distribution in the rail head versus depth caused by Hertz pressure acting on a circular area (a=6.8 mm, Fz =171.7 KN)
77
Figure 3.12 Octahedral shear stress ( ) of a rail subjected to a non-uniform (Hertzian) contact pressure (a = 6.8mm, Fz = 171.7 KN)
78
Figure 3.13 Longitudinal stress at the underhead offset of measurement point 20 mm
80
Figure 3.14 Rail section for residual stress distribution 82
List of Figures
xxii
Chapter 4
Figure 4.1 Longitudinal stress (S11) contour; (a) centric loading; (b) eccentric loading on gauge side (contact patch 30 mm offset); (c) eccentric loading on gauge side (contact patch offset of 30 mm with L/V=0.4 towards gauge side)
90
Figure 4.2 Longitudinal stress at the underhead offset of measurement point 34 mm; (a) at different contact patch offsets (0, 15 and 30 mm) with L/V=0; (b) at contact patch 30 mm offset with different L/V= (0, 0.2 and 0.4)
91
Figure 4.3 Longitudinal stress at different underhead offset of measurement points (20, 34 and 37 mm from longitudinal centerline of rail) at the contact patch 30 mm offset and L/V=0.4
92
Figure 4.4 Longitudinal stress distribution with depth on a vertical plane at 20 mm offset from the rail vertical centreline towards the gauge corner for the different contact patch offsets (0, 15, and 30) mm with L/V ratios (0, 0.4)
93
Figure 4.5 Longitudinal stress distribution with depth on different vertical planes at 16.5, 20, 25 and 35 mm offset from the rail vertical centreline towards the gauge corner at a contact path 30 mm offset and different L/V ratios (0, 0.2, and 0.4)
95
Figure 4.6 Longitudinal stress (S11) contours on the cross-section of rail with the contact patch offset of 30 mm at midpoint of rail span ( in longitudinal direction at x=2100 mm), during different lateral loads with the L/V ratio of 0, 0.2 and 0.4. The affected region at which the stress becomes tensile shifts up from the rail underhead radius to surface
96
Figure 4.7 Mechanical response at the underhead and base fillet with lateral shear tractions (inward and outward); (a) Longitudinal stress variations with different magnitude and direction of shear traction at the underhead offset of measurement point 34 mm along the midpoint (x = 2100 mm) of rail with contact patch 30 mm offset; (b) Inward lateral deformation (U2) contour with L/V = 0.4; (c) Outward lateral deformation (U2) contour with L/V = 0.4
97
Figure 4.8 Longitudinal stress variations at the underhead radius with vertical foundation stiffness at the contact patch 30 mm offset and L/V = 0.4: (a) along longitudinal position of the rail at the underhead offset of measurement point 34 mm; (b) at the underhead offset of measurement point 34 mm and at base fillet at midpoint of rail span (in longitudinal direction x=2100 mm) in between two middle sleepers
99
Figure 4.9 Longitudinal stress variations at the underhead radius with 101
List of Figures
xxiii
seasonal temperature at contact patch offset of 30 mm and L/V = 0.4 (a) along longitudinal position of the rail at the underhead offset of measurement point 34 mm; (b) at the underhead offset of measurement point 34 mm at midpoint of rail span (in longitudinal direction at x = 2100 mm) in between two middle sleepers
Chapter 5
Figure 5.1 Worn rail profiles [143] 107
Figure 5.2 The model description for the cross-section view of the worn rail (Head wear, HW = 25 mm) profile, the strain gauge measurement location and the elastic foundations with variations of contact patch size and shape
108
Figure 5.3 Variation of maximum longitudinal stress (S11) at the underhead region vs rail head wear for an eccentric load located at 20 mm offset from the rail centreline with L/V = 0.4
111
Figure 5.4 Longitudinal stresses (S11) distribution at the rail head vs the different head wear (HW) profiles for an eccentric load located at 20 mm offset from the rail centreline with L/V = 0.4
111
Figure 5.5 Longitudinal stress distribution at the underhead for the different contact locations vs the HW of 0, 22 mm and 25 mm
112
Figure 5.6 Variation of longitudinal stress at the gauge side underhead radius for eccentric loading from rail centreline towards the gauge side for different rail worn profiles versus the lateral traction coefficient
114
Figure 5.7 Longitudinal stress variations at the underhead radius with vertical foundation stiffness at the contact patch 20 mm offset and L/V = 0.4: at the underhead offset of measurement point 34 mm and at the base fillet at the midpoint of the rail span (in longitudinal direction x=2100 mm) in between two middle sleepers
115
Figure 5.8 Longitudinal stress distribution with depth on different vertical planes offset from the rail vertical centerline towards the gauge corner, at a contact patch offset of 20 mm L/V=0.4: (a) HW = 25 mm, (b) HW = 15 mm
117
Figure 5.9 Longitudinal stress distribution with depth on different vertical planes at 15, 20, 25 and 30 mm offset from the rail vertical centreline towards the gauge corner at a contact patch offset of 25 mm L/V=0.4 HW 20mm
118
Figure 5.10 Longitudinal stress distribution (S11) with depth on a vertical plane at a 25 mm offset from the rail vertical centerline towards the gauge corner for contact patch offsets of 20 mm with L/V 0.4
119
List of Figures
xxiv
Chapter 6
Figure 6.1 Stress analysis at underhead offset of 25mm for head wear of 22 mm, contact patch offset = 15mm, L/V=0.4
125
Figure 6.2 Representation of model setup for DV fatigue damage analysis 128
Figure 6.3 Description of UVARs 2 to 5 on a critical plane 131
Figure 6.4 INDENTEC-Hardness measurement apparatus (Courtesy of Swinburne University of Technology)
133
Figure 6.5 Rail sample specimen (section and the shape) for hardness testing 133
Figure 6.6 Hardness distribution along transverse plane (starting from underhead radius measurement point offset 20 mm) of high strength rail steels
134
Figure 6.7 The fatigue threshold stress is in a range of t-1 corresponding to failure at the t-1 for 106 stress cycles (the highest cycle) and at the at the DV magnitude of the fracture stress for 101 stress cycles (the lowest cycle)
135
Figure 6.8 Von Mises stress distribution at the rail underhead radius vs different head wear profiles: for eccentric loads located at different offset from the rail centerline with L/V = 0, 0.2, 0.4
137
Figure 6.9 Variation of fatigue damage with rail head wear at the underhead radius for an eccentric load located at (a) 15 mm offset from the rail centerline with L/V=0; (a) 20 mm offset from the rail centerline with L/V=0.2, for a plain C-Mn Head Hardened (HH) grade [80], with t-1 = 205 MPa
139
Figure 6.10 Variation of DV damage parameter versus rail head wear HW at the underhead radius for an eccentric load located at different offset from the rail centreline with L/V=0, 0.2, 0.4, for three rail grades: a plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat Treated (LAHT) grade [127], and Hypereutectoid (HE) Heat Treated grade [81]
141
Figure 6.11 DV damage parameter vs different contact patch offset with a load of L/V=0.2 for the HW of 18 and 22 mm
142
Figure 6.12 The damage accumulation via the head wear at the underhead radius for an eccentric load of L/V=0.2 at the contact patch offset of 20mm
143
Figure 6.13 Prediction of fatigue life under the extreme loading conditions 144
Chapter 7
Figure 7.1 Illustration of normal and tangential coordinates for a smooth crack [138]
150
List of Figures
xxv
Figure 7.2 Representation of a non-planar crack in three dimensions by two signed distance functions Ļ and Ļ [138]
151
Figure 7.3 A single crack at the gauge corner: (a) crack shape (crack length = R and surface length = 2R); (b) crack orientation ( = 70o); (c) fine mesh at highlighted enriched region including crack and contact patch potions; (d) Contact patch offset (CPO), L/V ratio and position of the crack from the centre of rail head cross-section
153
Figure 7.4 Loading steps at different positions relative to the crack 154
Figure 7.5 The maximum KI (KI max) at the crack front (R = 17 mm, = 90o) under the different loading positions
157
Figure 7.6 The maximum KI (KI max) at the crack front: as a function of crack length R (5, 17, 30) under the loads of L/V =0.2 at CPO = 20 mm
158
Figure 7.7 The maximum KI (KI max) at the crack front ( = 90o) under the loads of L/V =0.2 at CPO = 20 mm (a) no RS; (b) with RS
159
Figure 7.8 The SIFs at the crack front (R = 10 mm) under the local bending with L/V =0.2 at CPO = 20 mm for rail HW = 22 mm.
161
Figure 7.9 The SIFs at the crack front ( = 90o) under the local bending with L/V = 0.2 at CPO = 20 mm
163
Chapter 8
Figure 8.1 Appearance of RCF damage [33]: (a) longitudinal section with gauge corner cracking; (b) Transverse defect
166
Figure 8.2 The model description for multiple RCF cracks at the gauge corner
170
Figure 8.3 Development of fracture surface for Mode I, Mode II, Mode III and mixed-mode-loading of cracks [189, 190]
173
Figure 8.4 Variation of maximum (over the entire crack front position) SIFs during one wheel passage (loading cycle) with different contact patch offsets (CPO)
175
Figure 8.5 SIFs and Keq at the crack front when a wheel passes with CPO = 0 mm, L/V = 0 for crack (R = 10 mm, = 90o) to the rail running surface with HW 22 mm
177
Figure 8.6 SIFs and Keq at the crack front when a wheel passes with CPO = 15 mm, L/V = 0 for crack (R = 17 mm, = 90o) to the rail running surface with HW 15 mm
178
Figure 8.7 Figure 8.7 SIFs and Keq at the crack front when a wheel passes with CPO = 15 mm, L/V = 0 for crack (R = 10 mm, = 90o) to the rail running surface with HW 22 mm, for three rail grades: a plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat
Figure 8.8 SIF at the crack front when a wheel passes with CPO = 15 mm, L/V = 0.2 for crack (R = 17 mm, = 90o) to the rail running surface with HW 15 mm
182
Figure 8.9 SIF at the crack front when a wheel passes with CPO = 20 mm, L/V = 0.2 for crack (R = 10 mm, = 90o) to the rail running surface with HW 22 mm
183
Figure 8.10 SIF at the crack front when a wheel passes for crack angle = 90o to the rail running surface with different HW, crack sizes R, CPO and L/V, for three rail grades: a plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat Treated (LAHT) grade [127], and Hypereutectoid (HE) Heat Treated grade [81]
185
Figure 8.11 The effect of modelling seven multiple cracks on SIF predicted for the central crack with SIF on crack front when a wheel passes with CPO = 15 mm, L/V = 0 for crack angle 90o to the rail running surface with HW 22 mm
187
Figure 8.12 The effect of modeling multiple cracks on Keq SIF predicted for the central crack when a wheel passes with CPO = 15 mm, for crack angle 90o to the rail running surface with HW 22 mm; L/V 0, 0.2.
188
Appendices
Figure A.1 Dimension of generic rail section with a reference point set near the lower gauge corner (underhead radius position) [7]
222
Figure A.2 Eccentric vertical loading and lateral loading of rail [7, 15] 224
Figure B.1 Longitudinal stresses at UHR, 26 km, 41 km, 74 km - parent rail (gauge and field side) [32, 148]
Table 3.2 Results of simulation of rail deformation with respect to the support of elastic foundation
79
Table 3.3 Residual stress scatter in the worn rail section given above [136] 82
Chapter 4
Table 4.1 Effect of track support stiffness 100
Chapter 5
Table 5.1 Effect of stress state at the underhead region due to residual stress 110
Table 5.2 Effect of rail support stiffness on rail longitudinal UHR stresses (S11)
116
Chapter 6
Table 6.1 Material properties and constants of rail 129
Table 6.2 Definition of UVARs for the output variables 130
Table 6.3 Chemical composition [43, 130] 132
Table 6.4 The estimated fatigue limits at the UHR for different high strength rail grades
134
Table 6.5 Approximate comparison of predicted head wear limits for head hardened rail based on fatigue limits with extreme loading cases for different track conditions compared to current head wear limits [124, 148]
140
Chapter 7
Table 7.1 Crack model parameter values 155
List of Tables
xxviii
Chapter 8
Table 8.1 Crack model configurations simulated 172
Table 8.2 Predicted head wear limits based on fracture strength with extreme loading cases representative of different track conditions [valid for studied crack sizes (radius) R =10mm, 17 mm and 22 mm as shown in Fugures 8.2 and 8.10]
186
Appendices
Table B-1 Rail stress data - Longitudinal stresses at UHR, 26 km, 41 km and 74 km - parent rail (high gauge tension / compression) [32, 148]
231
List of Notations and Acronyms
xxix
List of Notations and Acronyms
a and b Major and minor semi-axes BEM Boundary element method C3D4 4 node 3D tetrahedral solid elements C3D8R 8 node 3D solid elements with reduced integration COD Crack opening displacement CPO Contact patch offset CWR Continuously welded rail C Constant (comparable to the fracture strain in monotonic loading) Crack growth rate DB Deutche Bahn DHH Deep Head Hardened rail grade Di Damage corresponding to the ith equivalent stress cycles DV Dang Van E Modulus of elasticity FEM Finite element method FZ, FY Normal and tangential forces Shear modulus GC Gauge corner GCC Gauge corner cracking HDPE High density polyethylene HE3 Hypereutectoid rail grade HH a plain C-Mn Head Hardened rail grade (AS 1085.1) HR High rail HW Head wear IHHA International heavy haul association IRT Institute of Railway Technology k Yield limit in cyclic shear K Stress intensity factor Keq Equivalent stress intensity Keq
max Maximum (over the entire crack front) equivalent stress intensity factor KI Mode I stress intensity factor on the crack front KII Mode II stress intensity factor on the crack front KIII Mode III stress intensity factor on the crack front KI
max Maximum (over the entire crack front) Mode I stress intensity factor KL Lateral (elastic foundation) stiffness/area
List of Notations and Acronyms
xxx
KV Vertical (elastic foundation) stiffness/area kĻ Torsional stiffness L Contact patch length (in section 2.2.2.1) L Lateral load L/V Ratio of lateral to vertical wheel load LAHT3 Low Alloy Heat Treated rail grade LEFM Linear Elastic Fracture Mechanics Ls Support length LSM Level set method M Multiple crack run number m+ Positive normal to the crack front MGT Million gross tonnes Mode 1 In-plane tensile opening mode Mode II In-plane shearing mode Mode III Out of plane tearing mode N Vertical applied load n+ Positive normal to the crack surface P Total load P(x) Normal pressure distribution P(x,y) Contact pressure PIRD Pilbra Iron Ore Railway Division Pm Mean pressure Po Maximum contact pressure q(x) Tangential pressure distribution R Crack length or crack radius RCF Rolling contact fatigue RDF Reverse detail fracture rn Rolling radius of the wheel Rrx Transverse radius of the rail RS Residual stresses Rwx Transverse radius of the wheel TSBC Transportation safety board of Canada s Crack separation S Rail span SMises Von Mises stress S11 Longitudinal stress SC Standard carbon SIF Stress intensity factor SWT Smith-Watson-Topper TDs Transverse defects u1 Longitudinal translation U2 Lateral deformation
List of Notations and Acronyms
xxxi
UHR Underhead radius uR1 Longitudinal rotation UVAR1 Equivalent stress EQ, (MPa) (The inequalityās Equation 6.1 left side) UVAR2 Angle between the x axis and the external normal, n, of the critical plane UVAR3 Angle Ī² between the y axis and the external normal, n, of the critical plane UVAR4 Angle between the y axis and the external normal, n, of the critical plane UVAR5 The shear stress at the critical plane, (MPa) UVAR6 The shear stress amplitude, a(t), (MPa) UVAR7 The damage parameter, Di, at the ith number of the equivalent stress cycle UVAR8 The predicted number of cycles to failure, Nf UVAR-M User output variable V Vertical load Poisson ratio VCCT Virtual crack closure technique WLRM Whole life rail model Ws Support width X Distance from the contact pressure peak to the crack mouth X-FEM Extended finite element method Crack orientation or inclination T Ultimate tensile stress Y Yield stress , Material parameters in Dang Van criterion Fatigue limit in alternate bending , Vertical and lateral bending inertia of the entire rail , Vertical and lateral bending inertia of the rail head only Shear yield stress Fracture toughness Threshold stress intensity factor Number of cycles to failure or crack initiation Number of cycles to failure by ratcheting Peak Hertzian pressure Fatigue limit in reversed pure torsion Strain to failure in a monotonic test , Axial fatigue ductility and axial fatigue strength coefficients /2 Maximum principal strain amplitude Longitudinal stress due to lateral head-on-web bending Longitudinal stress due to vertical head-on-web bending stress ( ) Time dependent hydrostatic stress Longitudinal stress due to lateral bending Thermal stresses Longitudinal stress due to vertical bending Warping stress
List of Notations and Acronyms
xxxii
Longitudinal bending stresses Maximum stress on the maximum principal strain plane The ith equivalent stress ( ) Time dependent shear stress amplitude on the shear plane Equivalent stress , Shear fatigue strength and shear fatigue ductility coefficients Octahedral shear stress Shear stress in a specified shear plane Closed cycle plastic strain range Ratchetting strain per cycle , Average ratchetting axial and shear strains per cycle Wear rate 2R Crack surface length Thermal expansion coefficient Traction coefficient Wheel point Ļ Crack surface, representative of crack kinking Ļ Orthogonal surface representative of crack twisting
Chapter 1. Introduction
1
Chapter 1
Introduction
1.1 Research background and motivation
Introduction 1.1.1
In order for rail operations to be safe and without interruptions, it is vital that the
railway industry maintains rails in the best possible condition. However, rail
maintenance is costly. For example, significant rail maintenance that includes the
reprofiling and repairing of rail defects costs the European Union several hundred
million Euros per annum. Recently, treatment of head checks alone was said to cost the
Netherlands around ā¬ 5 million per annum [11]. In the United States, the estimated
annual cost of replacing and repairing worn out and degraded rails is approximately US
$2 billion [12]. The railway companies in other countries spend similarly high amounts.
Two inevitable damage modes, head wear and rolling contact fatigue, are therefore of
major concern to rail maintenance and safety. Rails in high axle load conditions can
withstand severe head wear without increased safety risks. However, the effect of severe
head wear on RCF is not well understood and permitting heavy head wear either
through natural wear or after profile grinding to increase rail life may compromise rail
safety by increasing the risk of failure [1]. RCF is a primary damage mode in high
strength (Figure 1.1) and standard carbon (Figure 1.2) rail steels, occurring
predominantly at the top of the rail surface and in the gauge corner region. Additionally,
damage occurs at the lower gauge corner [6, 7] (Underhead radius, UHR).
Chapter 1. Introduction
2
Figure 1.1 Appearance of RCF damage on high rail [2]
RCF accounts for approximately 90% of all rail defects [13] such as head checks, gauge
corner cracking (Figure 1.1) and transverse defects (Figures 1.2 and 1.3). The
development of transverse defects (TDs) from RCF damage is considered a major
source of rail failure in parent rail under high axle load conditions [8, 13, 14], as shown
in Figure 1.2. Transverse defects, classified as detail fractures (Figure 1.3a) in North
American railroad tracks, account for around 75% of total rail failure in continuously
welded rails [8, 10, 15]. Transverse defects are considered to represent a potential risk
for rapid fracture, which can lead to catastrophic rail failure, as shown in Figure 1.3b [1,
2, 16, 17].
Factors contributing to the rapid fracture of long pre-existing RCF cracks, fatigue
damage initiation at the underhead radius (lower gauge corner) and, in particular, failure
risk associated with transverse defect (TD) development due to crack turning behaviour
(transition from Mode II/III to Mode I), is generally not fully understood for high axle
load conditions. Field investigations indicate that the occurrence of tension spikes at the
Eutectoid rail Hypereutectoid rail
Chapter 1. Introduction
3
rail underhead radius as a result of the lateral head bending of the whole rail profile, and
localised vertical and lateral bending of the head-on-web are significant if this
behaviour is to be understood.
Figure 1.2 Development of transverse defect [1]
Figure 1.3 Transverse defect (TD) developed from RCF damage [picture to the right is from [1, 2, 17], picture to the left is from [16]]
Standard carbon rail
(a) (b)
Chapter 1. Introduction
4
Application of a āWhole Life Rail Modelā to high axle load conditions 1.1.2
The most comprehensive methodology dealing with RCF cracks is related to the
Hatfield incident in the UK in the year 2000 [18-21]. The research work arising from
this event includes the development of a āWhole Life Rail Modelā āWLRMā (see Figure
1.4), which is used when designing maintenance strategies for a rail system (Kapoor et
al [18-20, 22-24], Burstow et al [30] and Dutton et al [21]). All phases of crack
development, from initiation to the growth created by contact stresses and from rail
bending stresses, were modelled in this approach. Figure 1.4a shows crack initiation and
early propagation by ratchetting (curve A). As the length increases the crack
propagation rate increases (curve B). However, relatively long cracks move away from
the contact stress field and the rate of propagation drops (curve C) until, finally, the
crack is driven by bending (curve D). The āWhole Life Rail Modelā also describes the
sequence of events in the development of RCF that leads to a rail break, as shown in
Figure 1.4b [22, 25, 26]. A rapid fracture resulting from a transverse defect (TD), such
as that shown in Figure 1.4b, may occur if an RCF crack turns down at a certain depth
from the rail head, also driven by tensile bending stress.
The longitudinal bending stresses included in the āWhole Life Rail Modelā considered
vertical wheel load, and as a result, the rail bends and a bending moment can be
generated [27]. With rail uplift ahead and behind the rail/wheel contact, longitudinal
tensile stresses develop due to rail vertical bending (reverse bending), as shown in
Figure 1.5. It is these stresses that cause the rolling contact fatigue cracks to turn down
into transverse defects (TDās) and make them grow until a final fracture takes place, as
has been reported in references [22, 23, 25-29].
The approach adopted in the āWhole Life Rail Modelā for the prediction of crack growth
behaviour appears relevant under high axle load conditions. By comparing the assumed
longitudinal bending response of a rail in the āWhole Life Rail Modelā with the in-track
measurements under high axle load conditions (Figure 1.6), it has been revealed that the
actual response of the rail section is more complex than that assumed in the āWhole Life
Rail Modelā. In the latter, there is a large and more variable tensile component when the
Chapter 1. Introduction
5
Figure 1.4 āWhole Life Rail Modelā āWLRMā (a):Crack propagation rate v/s length (adopted from [1, 20, 22-24, 30, 31]); (b): Sequence of events in the development of RCF [22, 25, 26]
(b)
(a)
Chapter 1. Introduction
6
wheel is directly on the top of the reference (measurement) position, as shown in Figure
1.6 [1]. This indicates that the effect of global rail bending stresses was considered in
the āWhole Life Rail Modelā but that the issue of local rail bending (lateral head
bending of the whole rail profile and localised vertical and lateral bending of the head-
on-web) was ignored. WLRM included the effect of discretised sleepers but the
dynamic and thermal effects were not considered
In addition to a whole rail cross-section vertically bending on an elastic foundation, the
rail head also undergoes lateral bending of the whole rail profile and vertical and lateral
bending of the head-on-web. This effect corresponds to an additional local bending
stress being superimposed due to the lateral bending of the whole rail profile and the
localised bending of the rail head-on-web. This results in a much greater magnitude of
tensile stresses (tension spikes, see Figure 1.6) at the underhead radius when the wheel
is directly on the top of the rail head than those associated with rail uplift ahead and
behind the contact patch (as was identified in the post-Hatfield study, see Figure 1.5).
The stress state generated near the underhead radius rapidly changes the stress
components where the geometry of the rail changes due to a sharp radius. It is in this
region that the longitudinal stress shows tension spikes when measured under high axle
loading conditions, as shown in Figure 1.6 [1, 32, 33].
The work presented here is the first step towards modelling the effect of lateral head
bending and localised head-on-web bending, in order to understand the possibility of the
failure risk associated with the rapid fracture behaviour of pre-existing RCF cracks and
fatigue damage initiation at the underhead radius. In particular, a qualitative assessment
of the tendency to form transverse defects due to Mode I transition in the presence of
tension spike at the underhead radius was conducted. In this thesis, underhead radius
stresses due to local bending were calculated for use as an input to fatigue life and rapid
fracture models.
Various researchers [1, 3, 4, 6, 7, 34-45] have analysed rail bending stresses with a
special focus on the additional local bending stresses superimposed due to the localised
bending of the rail head-on-web and lateral bending of the whole rail profile and will be
described next.
Chapter 1. Introduction
7
Figure 1.5 Assumed longitudinal bending response of rail in āWhole Life Rail Modelā [27]
Figure 1.6 Measured longitudinal stress response under in service loading; example shown for gauge side underhead radius of high rail in 609 m radius curve [1, 32, 33]
Tension spikes
Compressive stresses
Chapter 1. Introduction
8
Eisenmann [34] measured the tensile stresses at the underhead location in the field (at
lower gauge corner), and also calculated theoretically the tensile stresses at the gauge
and field side underhead location for eccentric and inclined loading. The local tensile
stresses on the field side underhead location due to lateral wheel loading were measured
under laboratory conditions and in the field by Sugiyama et al [35]. Marich [36-38] has
also reported that the presence of tensile longitudinal stresses at the underhead radius of
the rail are influenced by an increase in head wear.
Jeong [39], Jeong et al [7, 40], Orringer et al [46, 47] and Lyon et al [48] reported that
the rail head is supposed to be a beam resting on the rail web. The web is considered to
behave as an elastic foundation. Therefore, the local bending stresses are the result of
lateral bending of the whole rail profile and localised bending of the head-on-web.
These stresses are additional to the vertical bending of the whole rail profile (the so
called global bending). Orringer et al [47] analysed the longitudinal bending stresses in
the rail head in detail. They found it to consist of five possible components: (1) vertical
lateral head-on-web bending. The field measurements made by Orringer et al [47]
showed that, for tangent track, the longitudinal stress caused by rail head-on-web
bending is much smaller than that caused by rail vertical bending. The warping term of
the longitudinal stress is similarly small. On curved track, the additional components of
lateral head bending and lateral head-on-web bending significantly increase the
longitudinal stress.
Jeong [39], Jeong et al [7, 40] and Orringer et al [46, 47] also reported that bending
stresses can be calculated by assuming the rail to be a beam on a continuous elastic
foundation with additional localised bending response of the head-on-web, under
applied contact load and foundations, as shown in detail in Figure 1.7. Their model
includes the vertical (kv), the lateral (kL), and the torsional (kĻ) stiffness of the elastic
foundation of the whole rail profile, along with the rail head as a separate beam bending
on an elastic foundation formed by the web. Where, āeā and āfā are the locations of
vertical and lateral loads with reference to the shear centre of a rail, as shown in Figure
1.7. Orringer et al [47] also reported the complete live-load stress sequence of five
Chapter 1. Introduction
9
possible components that contributing to longitudinal bending stresses (see Figure
1.7), namely:
(1.1)
where
= Longitudinal bending stress component due to vertical bending (dominant
component due to the vertical wheel loading)
= Longitudinal bending stress component due to lateral bending
= Warping stress
= Longitudinal bending stress component due to vertical head-on-web bending
= Longitudinal bending stress component due to lateral head-on-web bending
Figure 1.7 Representation of localised head-on-web bending and lateral bending of the whole rail profile as a beam model on a continuous elastic foundation with eccentric vertical and outward lateral loading (adopted from Orringer et al [46, 47], Jeong et al [7], and Salehi [49])
šš„š„šµ
Head
Web
Base
Chapter 1. Introduction
10
The equations for rail bending stresses are given in Appendix A.
Due to the complete live-load stress sequence of five possible components contributing
to longitudinal bending stresses, , the underhead radius stresses become high for
the high axle loads that are common in heavy haul operations. This produces reverse
detail fractures (i.e. transverse defects that are initiated at the lower gauge corner of high
rail), as shown in Figure 1.8ā1.11. This has occurred in North American rail systems [3,
5-9] because the poorly lubricated, heavily worn curved rails on stiff tracks are
subjected to high axle loads. Failures due to fatigue cracking in the underhead radius of
aluminothermic welds have also been found in Australia on rails subjected to high axle
load conditions, as shown in Figure 1.12. In practice, fatigue failure at the lower gauge
corner is generally associated with the presence of pre-existing defects or stress
concentrations in the form of a sharp radius, which may include a cold rolled flow /
shear lip (as was evident in the reverse detail fracture, Figure 1.8-1.11 and complex
local geometry in rail welds.
Jeong [8] specifically mentions flow lips as initiators, in addition to the possible
contribution of higher residual stresses at rail welds (which has been confirmed in
subsequent research on aluminothermic welds by Salehi et al [42-45], and also Mutton
[50] in connection with aluminothermic welds). The reverse detail fracture failure mode
is influenced by longitudinal bending stresses at the underhead radius, as reported by
Jeong et al [6]. In the presence of heavily worn rail, the magnitude of these stresses
could be even higher. Therefore, a pre-existing defect when combined with severe head
wear should be considered as increased risk for rail safety and integrity. The local
bending stresses also superimpose on axial stresses. These include, but are not limited
to, the effect of track bed support stiffness, and residual and continuously welded rail
stresses. Additional stresses are associated with dynamic axle loads and cant deficiency,
and other stresses are produced due to the influence of several wheels.
This research adds to the current understanding of the contributing factors in the rapid
fracture of pre-existing RCF cracks and the initiation of fatigue damage at the rail
underhead radius, especially the qualitative assessment of the tendency to form
transverse defects due to Mode I transition. These investigations have been conducted
Chapter 1. Introduction
11
Figure 1.8 RDF development [16] Figure 1.9 RDF (Courtesy of TSBC [3, 9])
The equivalent modulus of elasticity and equivalent radius are calculated from
Equations 2.3 and 2.4 respectively. The equations for contact size, deformation and
maximum pressure as calculated by Johnson [58], are as follows
ā
(2.8)
ā
(2.9)
ā
(2.10)
More recent details of HCT, which include the subsequent relaxation in limitation and,
in particular, the extension of its application to practical situations, are given in [60, 61].
Wheel-rail contact pressure distribution 2.2.3
In a 3D wheel-rail contact problem, the more generic shape of a contact area on the
centre of the railhead surface is elliptical, having major and minor-semi axes of a and b
respectively. The wheel in contact with rail is shown in Figure 2.4. Iwincki [60]
reported governing equations to calculate the dimensions of the elliptical patch from the
curvature of the rail and wheel surface in the contact point, as given below:
.
/
| |
(2.11)
.
/
(2.12)
.
/
(2.13)
Where, N is the vertical load, Rwx and rn are the wheel transverse and rolling radii
respectively and Rrx is the transverse radius of the rail running surface, as shown in
Figure 2.4. The values m and n as a relation of are given in [60].
Chapter 2. Literature review
23
Figure 2.4 Rail-wheel Hertzian contact [60]
The ellipsoidal Hertzian pressure distribution is given as
( ) ā.
/ (2.14)
(2.15)
0
(
)
1
0
(
)
1
(2.16)
Where is the wheel load perpendicular to the contact patch, is the maximum
contact pressure. The stresses are presented in a cartesian coordinate system for the
wheel-rail contact case and will be discussed in section 2.2.4. Accordingly, x is along
āaā and y is along ābā respectively as shown in Figure 2.5.
Rwx
Rrx
rn
Chapter 2. Literature review
24
Figure 2.5 Hertzian pressure distribution at the contact patch [the picture to the left is from [62], whereas the picture to the left is from [63]]
2.2.3.1 Wheel conicity in the steering
In a curve, the leading wheelset tracks towards the high rail, outside of the curve,
following Newton's third law and the trailing wheelset will tend to roll towards the
inside, as shown in Figure 2.6a-b. As the wheels are conical the rolling radius of the
outer wheel is increased while on the inner wheel it is decreased, as in Figure 2.6b. Both
wheels are rotating at the same speed; the larger radius wheel tries to roll further than
the smaller radius wheel, thus steering the wheelset towards a radial alignment, when it
will roll smoothly around the curve. The opposite process happens on the trailing
wheelset as it moves inwards on the curve (Figure 2.6a). This provides creep forces to
yaw the wheelset self-steer relative to the rail. These forces are generated by the leading
wheelset moving out beyond the equilibrium rolling line to give an excess of rolling
radius difference. Similarly, the required steering forces at the trailing wheelset are
generated by moving inwards from the equilibrium rolling line [64].
Chapter 2. Literature review
25
Figure 2.6 (a) Vehicle on a curve; (b) Rolling radius difference [64]
On tangent track the contact normally occurs at the central region of the railhead cross-
section. However, even on tangent track flange contact can occur if the vehicle is
unstable or has hunting behaviour. Due to the profile design we also see the high rail
and outer wheel contact patch move toward the gauge face of the rail. And conversely,
the contact stays central or may move outward on the low rail. The contact patch
dimensions at the gauge corner would depend on wheel-rail profile match. The contact
patch dimensions in the lateral directions are greater as the contact patch moves towards
the gauge corner if wheel and rail profiles are reasonably conformal, which they tend to
be in heavy haul applications. The field reports suggest that conformal wear adapted
(nonstandard) profiles are commonly used in Australian heavy haul railway systems
[50]. The literature also suggest the dimensions of the major semi-axis along the
longitudinal direction are around 8-12 mm [65]. The wheel flange makes contact with
the gauge face and subjects the sides of the railhead to sliding contact which
corresponds to rather severe curving and in this case wear would most likely be the
main mode of deterioration.
(a)
(b)
Chapter 2. Literature review
26
Figure 2.7 Lateral shift of wheelset leads to contact patch offset from the center of railhead cross-section [66]
In all wheel-rail contact cases, especially on curved rail tracks, the contact patch
location changes as the vehicle tends to exhibit lateral shift. Additional lateral forces
Chapter 2. Literature review
27
may arise from cross-level errors or incorrect super-elevation (cant) in curved track. In
real life, the contact patch shape and size vary as the contact moves towards the gauge
corner. A detailed analysis of the lateral shift of wheel rail contact was presented by
Piotrowski et al [66], as shown in Figure 2.7. The lateral shift of a wheel set leads to a
contact patch offset from the centre of the railhead cross-section. The size of the contact
area is very small, typically 80-120 mm2, as stated by Marshall et al [67], approximately
the size of a thumbnail [29]. The shape of the contact area may be circular, elliptical or
even two ellipses (sometimes even joined together, as in Figure 2.7). The contact patch
size and location depends on the rail and wheel profiles and wheel-rail forces. Any
change in rail or wheel profiles (due to wear or due to profile grinding) will change the
contact patch size, shape and location (and thus the running band) [37]. Additionally,
the wheel-rail profile changes continuously due to on-going operational and
maintenance activities that ultimately affect the performance of the system.
Stress field under a Hertzian contact 2.2.4
In order to find Hertzian contact stresses a general representation of contact pressure
with a normal pressure profile p(x) on elastic half space is given in Figure 2.8. The
corresponding Hertzian stresses are represented in both the cartesian and polar
coordinate systems. The contact stresses due to Hertz pressure acting on a circular
contact area can be calculated as follows [58]. The calculations are for normal Hertz
pressure distribution without traction.
Chapter 2. Literature review
28
Figure 2.8 Representation of surface traction on elastic half space [58]
By using the equations 2.17-19, the analytical solution is calculated based on a circular
contact area = 145.3 mm2 for radius a = 6.8 mm, with Pm = 1182 MPa. The contact
stresses distribution versus depth caused by Hertz pressure acting on a circular area of
radius āaā is plotted in Figure 2.9a.
Similarly the contact stresses as a result of Hertzian pressure distribution of contact
between two cylinders is given by Johnson [58]. The following equations are used for
the calculations of subsurface stresses along the axis of symmetry and are plotted with
ratio of stresses to maximum contact pressure as shown in Figure 2.9b. The
maximum shear is plotted in Figure 2.9c.
2 .
/ 3 (2.20)
.
/ (2.21)
Chapter 2. Literature review
29
[*( ) + ( )] (2.22)
The signs of m and n are the same as the signs of z and x respectively.
Figure 2.9: (a) Stress distribution versus depth caused by Hertz pressure acting on a circular area of radius āaā [58] (b) subsurface stresses for line contact; (c) contours of principal shear stress for line contact [58]
Stresses for more general cases of contact pressure distribution are provided in ref [58]
2.3 Rail failure mechanism
Different regions of the rail are affected differently due to induced stresses and
deformations as a result of rail operations. In the region close to the contact zone, the
plastic deformation due to the repeated rolling / sliding contact between the wheels and
the rail is of great significance. This plastic deformation causes wear, crack initiation
and propagation leading to rail defects such as head checks, gauge corner cracking and
squats if the contact loads are sufficiently high compared to the strength of the rail
material(s). Several models exist to predict the initiation and propagation phases of
crack development. The shakedown limit approach and a more scientifically advanced
0
0.5
1
1.5
2
2.5
3
-1.5-1-0.50
Dim
ensi
onle
ss d
epth
(z/
a)
Ratio of stresses to mean pressure ( )
(a) (c) (b)
Chapter 2. Literature review
30
methodology, āWhole Life Rail Modelā (WLRM), utilised in the post-Hatfield study, are
the most extensively used approaches. The WLRM was developed through extensive
scientific research activities combined with both numerical modeling and field
correlations. A description of the application of WLRM will be presented in section 2.7.
Several recent studies suggest the use of the shakedown limit approach [68-71] for
describing crack initiation phenomena.
Plastic flow and shakedown 2.3.1
In wheel-rail contact the material is subjected to repeated loading. How a material
responds to repeated loading depends on the magnitude of applied load and this can be
used to explain the concept of a shakedown limit. If the maximum stress is below the
elastic limit, the material will behave in a perfectly elastic manner, as shown in Figure
2.10. However, when the maximum stress is greater than the elastic limit, plastic
deformation occurs. The deformation results in residual stresses when the material is
unloaded. Strain hardening increases the (apparent) yield limit and the steady state
behaviour is perfectly elastic. This is called the elastic shakedown limit, below which
failure is due to high cycle fatigue. This leads to the long life of a component [72, 73].
Figure 2.10 Material response to repeated loading [73]
Chapter 2. Literature review
31
Johnson [74, 75] and Bower et al [76] developed a shakedown map, as shown in Figure
2.11. The traction coefficient ( ) levels are plotted against the x-axis, and the ratio of
the maximum contact pressure, Po to yield limit in cyclic shear, k, of the material is
plotted along the y-axis. The corresponding elastic, elastic-perfectly plastic, and
kinematic hardening shakedown limits are plotted in Figure 2.11.
Bower and Johnson [76, 77] also presented shakedown limits for a line and point
contact with longitudinal and lateral tractions. The relationship for the shakedown limit
of line contact is given below [76]:
{
.
/
}
(2.23)
Where
and .
/
Figure 2.11 Shakedown map for repeated sliding of a rigid cylinder over an elastic-plastic half space [74, 75, 77]
Chapter 2. Literature review
32
Low cycle fatigue and ratchetting failure 2.3.2
When the contact pressure is above the elastic shakedown limit and below the plastic shakedown limit, no accumulation of plastic strain occurs and the phenomenon is called cyclic plasticity with a closed cycle, as shown in Figure 2.10, the cyclic plastic shakedown limit or ratchetting threshold. The failure of material is generally by low cycle fatigue (LCF). Failure by low cycle fatigue life was studied by Coffin-Manson, and the equation given, as reported by Kapoor [72], is:
(
)
(2.24)
Where is the number of cycles to failure by low cycle fatigue, is the closed cycle
plastic strain range, and is constant with a magnitude comparable to the fracture strain
in monotonic loading, and n is equal to 0.5.
If the load exceeds over the plastic shakedown limit, ratchetting occurs, as shown in
Figure 2.10. There is an open elastic-plastic loop, and the material accumulated uni-
directional plastic strain in each cycle though the ratchetting process. For example, the
cross-section of material subjected to rolling contact reveals large unidirectional shear
plastic strain accumulation (>10) near the surface due to ratcheting. This is possible
because of high hydrostatic compression of about 1.5 GPa. The ratchetting failure
occurs if the accumulated plastic strain reaches the critical strain to failure, as the
ductility of the material is exhausted. This failure of surface material leads to cracking
and wear [72, 78].
Kapoor [72] reported that the number of cycles to failure due to ratchetting is as follows
(2.25)
ā( ) ( ā ) (2.26)
Where is the number of cycles to failure by ratchetting, is the ratchetting strain
per cycle expressed in terms of and , which are the average ratcheting axial and
Chapter 2. Literature review
33
shear strains per cycle as given in equation 2.26. Kapoor [72] hypothesized that low
cycle fatigue and ratcheting failure are competitive and the criterion in Equation 2.26
should be compared with low cycle fatigue crack initiation criteria. The one that has the
lower number of cycles to failure is activated.
For high axle load conditions, Welsby et al [79] used the shakedown diagram to
illustrate the relative behaviour of standard carbon and high strength rail steels in heavy
haul applications (Figure 2.12).
Figure 2.12 Shakedown diagram and RCF predictions [79] (high rail on a moderate curved track with axle load of 30 tonne)
The latter included plain C-Mn head hardened (HH) grade [80] and hypereutectoid (HE)
heat treated grade [81]. The shakedown map suggests ratchetting is expected in standard
carbon as the kinematic hardening shakedown limit is exceeded for a high rail on a
moderate curved track under an axle load of 30 tonne. Repeated plastic deformation will
lead to surface and subsurface plastic flow and damage, as was discussed in the studies
mentioned above. In the case of a plain C-Mn head hardened (HH) grade, the
Chapter 2. Literature review
34
shakedown ratio decreases within the elastic shakedown region. After some initial
yielding, the material stabilises and hence behaves more-or-less elastically under
subsequent similar loading. Some initial mild deformation can be expected under these
conditions. No accumulation of surface and subsurface damage is expected. For
hypereutectoid (HE) heat treated grade, the shakedown ratio is below the elastic limit
and the material will behave in a perfectly elastic manner and no deformation is
expected under these loading conditions [79].
2.4 Wear and rolling contact fatigue
Many researchers have extensively studied wear and RCF formation in wheel - rail
contact [48, 82-85]. Rail can fail by either wear (loss of profile) or by RCF (long
surface cracks or complete rail break) and requires replacing. This behaviour is of great
significance to the rail industry due to the increased risk of rail failure. A detailed
description of wear and RCF is provided next.
Wear 2.4.1
Wear is defined as damage to one or both surfaces as a result of the progressive loss of
material due to relative movement [60, 86]. Williams [87] has conventionally quantified
wear rate as the volume lost per unit sliding distance i.e. m3 / m = m2. The loss is tiny
but leads to a loss of functionality of the component. Economic consequences of wear
are widespread and pervasive. The wear rate can be calculated according to Archard
wear equation as:
(2.27)
Where is the wear rate, which is directly proportional to the load on the contact
patch P and dimensionless coefficient K, is inversely proportional to material
hardness H. The literature survey on wear fatigue interaction and allowable wear limits
is given in sections 2.7 and 2.8.
Chapter 2. Literature review
35
Rolling contact fatigue 2.4.2
Rolling contact fatigue is the process of crack initiation and propagation due to the
stresses caused by rolling/sliding contact. RCF occurs in ball bearings, gears, wheel -
rail, and many other components where components roll or slide. Wheel - rail material
develops small crack-like flaws as a result of many passes of the wheel. These crack-
like flaws grow into cracks due to repeated loading by a rolling/sliding contact and may
develop into rail breaks [29]. RCF damage occurs, predominantly at the top of the rail
surface and gauge corner regions (see Figures 1.1-1.3). Additionally, damage occurs at
the lower gauge corner (Underhead radius, UHR), as has already been shown in Figures
1.8-1.12.
Figure 2.13 Photomicrograph of an etched sectioned test sample from twin disc test [88]
Fletcher el al [88] presented a computer simulation of wear and rolling contact fatigue
considering surface roughness and additional failure mechanisms, and includes the
integration of crack initiation with the wear simulation as shown in Figure 2.13. They
Chapter 2. Literature review
36
investigated RCF behaviour in terms of cyclic ratchetting material response and wear
simulation, based on a twin disc test.
Surface and subsurface rolling contact fatigue 2.4.3
RCF cracks are generally classified as surface and sub-surface initiated. Typical surface
initiated cracks are found in modern rails as head checks [2, 25, 29, 63, 89], which are
clusters of fine surface cracks with an interspacing of approximately 0.5-10 mm at the
gauge side of high rails [90], as shown in Figure 1.1 [1, 2]. These surface initiated
rolling contact fatigue (RCF) cracks may turn downwards vertically through the rail
head as transverse defects (TD) (Figure 1.2), as examined by Mutton et al [1, 2] under
heavy haul conditions. It was reported that the presence of a single or multiple TDs in
heavily worn rail increases the derailment risk [1].
Figure 2.14 Schematic of the mechanism of RCF crack formation and branching [91]
Datsyshyn et al [91] has also reported the mechanism of RCF crack formation and
branching. A squat type defect was investigated in these studies. Rail bending, the
friction force, the thermal and residual stresses and hydraulic effects were identified as
major contributing factors for RCF crack branching, leading to transverse defects, as
shown in Figure 2.14 in addition to material anisotropy and texture. It was not
Chapter 2. Literature review
37
confirmed which of these factors or combination leads to turning down as a transverse
defect.
Jeong [8, 10] contended that a transverse defect originating close to the running surface
of the rail head, is classified as a detail fracture (DF) by the Federal Railroad
Administration in North America, as given in Figure 2.15 [8]. Jeong [8, 10] defined the
detail fracture as, ā a progressive fracture starting from a longitudinal separation close
to the running surface of the rail head, then turning downward to form a transverse
separation substantially at right angle to the running surfaceā. A detail fracture is
associated with RCF damage and is classified as a transverse defect (TD) in Australian
railway systems. These are two different names for the same phenomenon in two
different systems. Jeong [8, 10, 15] reported that, based on railroad safety statistics from
the US Department of Transportation, Federal Railroad Administration (FRA), detail
fractures account for approximately 75% of the rail defect population in continuous
welded rail (CWR) tracks in North America. Transverse defects and detail fractures
arise as a result of both surface and subsurface initiated cracking.
Figure 2.15 Broken rail showing shell origin of detail fracture [Courtesy of Transportation Technology Centre, Inc. (TTCI)] [8, 10]
Chapter 2. Literature review
38
Shells are subsurface initiated cracks, particularly the initiation of rolling contact fatigue
cracks at the inclusion bands that are normally present in rail steels that were
manufactured prior to the introduction of improved steelmaking technologies such as
vacuum degassing and continuous casting. Shells can generally be more shallowly and
deeply initiated and different load components are dominant causes for the different
phenomena of surface and subsurface initiation. Shells normally turn down to give rise
to detail facture. Figure 2.15 reveals a typical broken rail showing the shell origin of a
detail fracture [8, 10].
Bower et al [76] reported that the rail surface layer undergoes plastic deformation to a
depth of around 15 mm. Clayton [92] presented a review of experimental research based
on tribological aspects of wheel-rail contact. It was reported that the nucleation of
fatigue cracks is deeper, at around 3-15 mm from the rail head running surface. The
crack nucleation at these positions was attributed to the presence of large tensile stresses
[93, 94]. Sugino et al [93] suggested that shells may occur in the absence of inclusions.
Farris [94] argued that subsurface shells are RCF cracks initiated at inclusion in older
rail steels. The material defects associated with geometric conditions such as flow or
shear lip formation in the case of RDF in parent rail and stress concentration at the
collar edge of the rail weld which give rise to stress concentration. These are the most
common reasons for the initiation of subsurface RCF cracks in high rails under high
axle load conditions.
Rail underhead radius failure 2.4.4
The underhead (UHR) region is away from the contact surface but is influenced by local
bending behaviour due to wheel-rail contact conditions. When the head wear increases,
under the same wheel-rail contact conditions, stresses in the underhead region become
critical. Damage investigation conducted by Canadian Pacific Railway [9] concluded
that the response of the underhead region may also be relevant to the propagation
behaviour of defects, which are initiated at, or close to, the primary wheel-rail contact
region. It is within this region, which is subjected to high contact stress and creepage
levels, that gauge corner collapse, plastic flow, shelling, and the development of
Chapter 2. Literature review
39
transverse defects may occur [9]. In sharper curves, as the curving forces increase, the
rail material potentially results in a flow lip at the lower gauge corner (underhead
radius). With the continued passage of rail traffic, a fatigue defect may be initiated and
propagated into the rail head [9].
Jeong et al [7] studied the influence of longitudinal bending stresses at the underhead
radius (lower gauge corner of rail) on the behaviour of reverse detail fractures (i.e.
transverse defects which initiate at the lower gauge corner of heavily-worn rail). This
defect type, observed in poorly lubricated, heavily worn curved rails on stiff track
subjected to high axle load conditions in the North American rail systems [6, 7, 9] was
shown in Figures 1.8-1.11. A similar failure mode due to the initiation of fatigue
damage at the underhead radius at aluminothermic welds in heavy axle load railway
operation in Australia is shown in Figure 1.12. However, such welds normally exhibit
lower material strength and higher residual stress levels than in normal rail, in addition
to the complex geometry associated with the presence of a weld collar, leading to stress
concentration. A reverse detail fracture is occasionally seen in aluminothermic welds
but generally occurs in the rail underhead [8]. Jeong specifically mentions flow lips as
initiators, in addition to the possible contribution of higher residual stresses at rail welds
(which has been confirmed in subsequent research on aluminothermic welds by Salehi
et al [42-45], and also by Mutton [50] in connection with aluminothermic welds).
Marich [36-38] reported the presence of high tensile longitudinal stresses at the
underhead radius (lower gauge corner) of the rail due to localised head-on-web bending.
The tensile stress was a function of vertical and lateral load eccentricity and was
influenced by changes in the head wear. A comparison of different worn profiles of 60
kg/m and 68 kg/m rail sections was conducted under eccentric vertical and lateral
loading. The FE results revealed that a smaller rail section would be preferable in terms
of allowable rail head wear due to the presence of lower stress values [36], as shown in
Figure 2.16. This did not take into consideration the increased beam strength with the
larger rail, which may be necessary for the axle loads considered. This work also did not
take into consideration the possible impact of RCF damage. Subsequent research by
Chapter 2. Literature review
40
Mutton et al [1] demonstrated that the approach taken by Marich was therefore non-
conservative in the presence of RCF damage.
Figure 2.16 Predicted head stresses in 60 kg/m and 68 kg/m rails [36]
The results of the analysis reported by Marich [37] demonstrated that the longitudinal
stresses at the underhead radius are increased by moving the contact patch away from
the centre of the rail head cross-section on a heavily worn rail, as shown in Figure 2.17.
An increase in axle load from 30 to 35 tonnes, along with a dynamic load factor of 1.5,
results in a corresponding increase in longitudinal stress, as shown in Figure 2.17.
Marich [36] also proposed modifications in geometry at both the gauge and field side
underhead radius positions. The results of these modifications are presented in Figure
2.18. The plots suggest that the current rail sections are far from optimum in terms of
assessment for critical rails stresses. The literature suggests that very little work has
been conducted regarding the modifications proposed at the underhead radius.
0
100
200
300
400
500
600
700
0 5 10 15 20 25 30 35 40 45 50 55 60
Long
itudi
nal s
tres
s (M
Pa)
Percentage rail head loss
68 kg/m, @ 22.5 mm eccentricity60 kg/m, @ 22.5 mm eccentricity68 kg/m, @ 15 mm eccentricity60 kg/m, @ 15 mm eccentricity
Chapter 2. Literature review
41
Figure 2.17 Effect of contact patch offset (CPO) on longitudinal stresses at underhead radius (UHR) for 68 Kg/m rail at nominal axle load of 30 and 35 tonnes, dynamic factor of 1.5 and L/V ratio of 0.2 [37]
Figure 2.18 Predicted rail head stresses for different head designs [36]
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25 30 35 40 45 50 55 60
Long
itudi
nal s
tres
s (M
Pa)
Percentage rail head loss
Dynamic wheel load 220.50 kN, @ 15 mm eccentricity
Dynamic wheel load 220.50 kN, @ 22.5 mm eccentricity
Dynamic wheel load 257.25 kN, @ 15 mm eccentricity
Dynamic wheel load 257.25 kN, @ 22.5 mm eccentricity
0
50
100
150
200
250
300
350
400
0 5 10 15 20 25
Long
itudi
nal s
tres
ses (
MPa
)
Eccentricity of vertical load (mm)
Modification 2Modification 1Standard
347 MPa, Stress limit for SC rails
Chapter 2. Literature review
42
Crack initiation and propagation is of great significance to the rail industry due to the
increased risk of rail failure, as described above. Predictions associated with this
behaviour are based on analytical, finite element analysis and experimental techniques.
However, analytical and experimental techniques are beyond the scope of this thesis.
Finite element analysis techniques are reviewed below. Several models exist to predict
the initiation and propagation phases of crack development. The literature suggests
many approaches and in the following sections the important models for crack initiation
and propagation are described.
2.5 Crack initiation
As described above, in a rail section, the contact and underhead radius regions are two deformation zones that can be simply classified in terms of the mechanical responses caused by the wheel-rail contact conditions. The former region, which is close to the contact surface, is the region of plastic deformation due to repeated rolling/sliding contacts between the wheels and the rail. As a consequence, if the contact loads are high, a crack initiates and propagates in the contact region leading to rail defects such as rolling contact fatigue. This behaviour is of great significance to the rail industry due to the increased risk of rail failure. The possibility of catastrophic rail failure and consequent fatalities has resulted in a great deal of research [48, 82-85, 95-98]. The underhead radius (UHR), which is away from the contact surface but is influenced by local bending behaviour due to wheel-rail contact conditions, results in rail defects such as reverse detail fractures (RDF), as was described in section 2.4.4. As stated above, the underhead radius (UHR) becomes critical when the wear increases in the rail head under the same wheel-rail contact conditions.
A study of the cyclic ratcheting material response and the RCF of a pearlitic rail steel was conducted by Ringsberg et al [68], based on Dang Van criterion, Coffin-Manson and SWT criterion and ratchetting failure. Dang Van criterion will be described in the next section 2.5.2. In Coffin-Manson, the failure of material is due to low cycle fatigue. For shear dominated failure, this is based on maximum shear strain amplitude, as given in Equation 2.28.
Chapter 2. Literature review
43
( )
( ) (2.28)
Where and are the shear fatigue strength and the shear fatigue ductility coefficient, b and c are the fatigue strength and the fatigue ductility exponents, G is the shear modulus and is the number of cycles for crack initiation.
The SWT is an energy based criterion and is based on mean stress, as given in Equation 2.29
. /
( )
( )
(2.29)
In Equation 2.29,
is the maximum principal strain amplitude and is the
maximum stress on the maximum principal strain plane. In addition and are the
axial fatigue ductility and axial fatigue strength coefficients, b and c are the fatigue
strength and the fatigue ductility exponents. The approach given in Equation 2.28 by
Coffin-Manson underestimated the crack initiation, whereas the SWT approach
(Equation 2.29) overestimated it but later predicted better results
For cyclic ratchetting material responses, Ringsberg et al used finite element (FE) simulations for different magnitudes of contact pressure and traction forces in the rolling contact. A large accumulation of shear deformation of material was found at the sub-surface of the running band. The ratchetting material model calculates the accumulation of shear plastic strains on the contact surface under different contact pressures and traction forces. RCF damage develops when the accumulated strain reaches the critical value for the rail material [68]. Two rail materials have been considered in order to prevent the ratchetting related failure described by Ringsberg et al [70]. They used functionally-graded material, i.e. the application of a surface coating. This is an option, but it is unclear if this would prevent ratchetting failure under all conditions.
Chapter 2. Literature review
44
Multi-axial fatigue criteria 2.5.1
A number of methods have been proposed for the prediction of crack initiation based on multi-axial fatigue criteria. The hypothesis for these studies is generally based on the theory of plasticity with a suitable fatigue criterion for rail life prediction. Equivalent stress criterion proposed by Dang Van [28, 53, 99-103] and a more complex energy-density based model proposed by Jiang and Sehitoglu [104] are the most extensively used approaches.
To predict fatigue crack initiation, a comparison of multi-axial fatigue criteria as applied to rolling contact fatigue was conducted by Conrado et al [105] and Ciavarella et al [106]. Ciavarella et al [106] compared the Dang Van, the Crossland and the Popadopoulos criteria and results of the comparison were obtained. Conrado et al [105] investigated the comparison between the Dang Van and the Liu-Zenner criterion for an assessment of fatigue limits of wheel-rail contact. Significant differences of predicted contact fatigue limits were obtained. It was suggested that careful correlation with experimental data is needed for the validity of predicted results using these criteria.
Ekberg et al [107, 108] developed an engineering model for fatigue damage prediction
in railway wheels. A comparison of sub-surface and surface initiated fatigue damage
was conducted based on the Dang Van and the shakedown model respectively, to
determine the dominating mechanism. The strengths and weaknesses of each model for
wheel fatigue damage analysis were elaborated. In another study, Ekberg et al [109]
reviewed multi-axial fatigue criteria as applied to rolling contact fatigue for the Sines,
the Crossland and the Dang Van criteria and a comparison of these criteria was
presented based on the strength and weaknesses of each model. As the wheel-rail
contact loads cause a multi-axial state of stress with out-of-phase stress components and
varying principal stress directions, it was suggested that the Dang Van criterion would
be suitable under these circumstances, as compared to the Sines and the Crossland
criteria and was thus required for predicting fatigue crack initiation and damage life in a
rail for this study. Ekberg et al [62] also used a high cycle fatigue model based on the
Dang Van theory for fatigue life and damage prediction in wheel ā rail rolling contact.
The Dang Van fatigue initiation criterion has previously been used in many industrial
Chapter 2. Literature review
45
applications, particularly in the automotive industry [28, 52]. The fatigue criterion is
based on a multi-scale approach and on a shakedown limit hypothesis.
Dang Van criterion 2.5.2
The Dang Van (DV) [28, 51, 53, 99-103] criterion is a shear stress based criterion,
which is applicable for stress levels below the elastic shakedown limit of the material. If
the following inequality is satisfied on a shear plane passing through each material point
at least once in the whole stress cycle, damage occurs. This inequality is expressed as:
( ) ( ) (2.30)
The value of the inequalityās left side represents a numerical index for fatigue damage.
Ļa(t) is the time dependent value of shear stress on the specified shear plane at the
specified material point and is defined as the difference between the instantaneous and
mean shear stress of the loading cycle; Ļh is the time dependent hydrostatic stress at the
material point. The constants (aDV and bDV) are the functions of material fatigue limits
[28]. The fatigue limits, f-1 and t-1 can be obtained from the classic experimental bending
and twisting tests, respectively. The constants are calculated as:
( )
(2.31)
As reported by Dang Van et al [52, 53, 99], if the rail material fatigue limits are, f-1 =
470 MPa and t-1 = 270 MPa, then the fatigue prediction loading path at different points
of the rail surface suggests fatigue failure may be initiated in the central zone of the rail
head surface due to the presence of shear traction as shown in Figure 2.19.and that
ought to depend heavily on contact geometry etc,
Chapter 2. Literature review
46
Figure 2.19 Fatigue prediction loading paths at different points of the rail surface [52, 53, 99]
Critical shear plane 2.5.3
The plane on which the mentioned inequality (Equation 2.30) is satisfied is called the
critical plane. However, the critical plane is not obvious at the beginning of the analysis,
so the inequality should be assessed in all shear planes passing through each material
point being investigated for potential fatigue crack initiation. A description of critical
plane with shear stress, at a material point āOā in a shear plane defined by normal
vector ānā, with respect to coordinate system āxyzā (adopted from Ekberg et al [62]) is
given in Figure 2.20. The stress history during a stress cycle can be evaluated for
comparison to the fatigue threshold limits of the material. The Dang Van critical plane
approach is advantageous, compared to the Sines or Crossland criteria [109]. Ekberg
[103] also employed another approach based on the amplitude of mean shear stress,
defined as the centre of the minimum circle circumscribing the tip of the shear stress
vector of the loading cycle. The aforementioned approach was adapted in this thesis.
Chapter 2. Literature review
47
Figure 2.20 Representation of critical plane with shear stress at a material point in a shear plane defined by normal vector ānā, coordinate system xyz (adopted from Ekberg et al [62])
Damage Accumulation 2.5.4
To quantify the damage, the Palmgren-Miner linear damage accumulation rule [103], in
conjunction with the Wƶhler curve [62], was also used. In terms of the Palmgren-Miner
linear damage accumulation rule, the damage degradation can be determined through
the following expression:
Di = ( ) ( ) (2.32)
Where, Di: is the damage corresponding to the ith equivalent stress cycles, i: is the
number of the shear stress cycles, amd ĻEQi: is the ith equivalent shear stress cycles
calculated by Eq. (2.30). The damage degradation at each material point via the ith stress
cycles to failure, Nf, can be defined as:
Di = 1/Nf. (2.33)
O
Chapter 2. Literature review
48
2.6 Linear elastic fracture mechanics
Linear elastic fracture mechanics (LEFM) are based on the stress intensity factor (SIF).
The stress intensity factor is used to quantify the stress state (āstress intensityā) at the
crack tip caused by multi-axial loading that is either remotely applied or is residual
stress. The stress intensity factor (K) is defined as the product of the nominal stresses in
a body, and the square root of the half-length of the crack (a), as given in Equation 2.34
ā (2.34)
Figure 2.21 Modes of crack growth (a) Mode I, opening (b) Mode II, shearing (c) Mode III, tearing [20]
Where Y is a factor depending on the geometry and location of the crack and the
loading conditions. K can be used to predict fracture, but to predict crack growth , ĪK is
needed. The crack will grow when a crack stress intensity factor range (as a result of
repeated loading of rail) exceeds an experimentally determined threshold. There are
three modes of growth, categorised as Mode I (opening), Mode II (shearing) and Mode
III (tearing), as shown in Figure 2.21. The calculated stress intensities are converted into
a crack growth rate by using a crack growth law [20, 110-113].
Chapter 2. Literature review
49
Crack propagation and rapid fracture 2.6.1
The possibility of crack propagation and the rapid fracture of pre-existing RCF under
Mode I loading has been investigated by a number of researchers. Most of the attempted
crack growth studies use finite element analysis procedures. In a study to develop a
strategy for the life prediction of RCF crack initiation and propagation, Ringsberg [114]
conducted multi-axial fatigue analysis to predict the fatigue crack initiation and
propagation life of RCF cracks, as shown in Figure 2.22. Elastic plastic finite element
modeling was conducted, along with multi-axial fatigue crack initiation analysis, using
a critical plane approach. These investigations were correlated by laboratory and field
measurements.
Figure 2.22 Three phases of the life of a RCF crack initiated at the surface [95, 114]
In another study, Ringsberg et al [95] described the crack propagation stages in which
small cracks with a length of around 0.1 mm initially grew at a shallow angle of 10-25o
until they reached a critical length of 1-2 mm. At this length the fatigue parameter as a
weighted sum of products of stress and strain at the crack tip will govern the crack
propagation either upward or downward to the rail surface, as shown in Figure 2.22.
The same behaviour of crack branching was reported by Wong et al [115]. Fischer et al
[116] investigated crack propagation behaviour influenced by the loading and material
Chapter 2. Literature review
50
characteristics of rail steel. Initiation and propagation of cracks in rail steel are
unavoidable due to rail operations. The qualitative result suggested that the fatigue
resistance of a rail against crack propagation could be improved by variations in
strength of the rail materials. Fischer et al [116] also reported that a crack branches
downward leading to a break in the rail if high tensile stresses due to low temperature
are present and the Mode I stress intensity factor exceeds the threshold.
Figure 2.23 Fluid assisted mechanisms of crack growth: (a) shear mode crack growth, accelerated by fluid reduction of friction between the crack faces, (b) hydraulic transmission of contact pressure, (c) entrapment and pressurization of fluid inside the crack and (d) squeeze film pressure generation [117, 118].
Fletcher et al [117, 118] investigated the effect of fluid penetration in RCF cracks that
leads to modification of the crack face friction and ultimately affects and accelerates the
Mode II growth of RCF cracks. Figure 2.23 presents shear mode growth influenced by
friction between crack faces, and shear mode crack growth that is accelerated by fluid
due to a reduction of friction between the crack faces. Other fluid assisted mechanisms
Chapter 2. Literature review
51
of crack growth, such as hydraulic transmission of contact pressure, entrapment and
pressurization of fluid inside the crack and squeeze film pressure generation are also
shown in Figure 2.23.
Fletcher et al [119] presented semi-elliptical crack modelling using a boundary element
approach. The contact load was included along with variable contact patch shapes
passing on different crack shapes and orientations. Greenās function was used to
quantify the SIF at the crack faces. Kapoor et al [18] conducted an examination of the
influence of residual stress distribution on the growth of RCF cracks. During crack
closure, the residual stresses can increase the stress intensity factor and change the
loading cycle. It was found that the compressive residual stresses in the vertical
direction were more significant than the longitudinal direction, as they suppress the
tendency for RCF cracks to develop Mode I growth [18].
Liu et al [120] conducted crack propagation analysis based on a critical plane approach
for a semi elliptical crack in a wheel representative of subsurface initiated cracking. As
the wheel is subjected to the same contact pressure, a subsurface crack in the wheel is
therefore comparable with a RCF crack in the rail. Additionally, wheels may contain
different initial residual stress distributions, due to differences in the manufacturing
processes between wheels and rails. In this study both the crack propagation direction
and the growth rate were investigated. The results suggested that a crack propagates
faster in a semi major axis direction as compared to semi minor axis direction.
Bogdanski et al [121] studied crack propagation by considering the stress intensity
factor influenced by crack pressurization due to water entrapment. A semi elliptical
crack was considered, located on the surface of a block with fixed boundary conditions.
The results suggested that the entrapped water increased the opening mode stress
intensity factor resulting in an increase in the crack growth rate.
Farjoo et al [122] conducted 2D crack modelling on the surface of a rail laid on 9
sleepers, using FEM and X-FEM modelling approaches that incorporate the elastic
foundation effects. The sleepers and ballast were modelled as springs. It was found that
stresses due to the elastic foundation increased the crack growth rate. Farjoo et al [123]
Chapter 2. Literature review
52
also conducted a 3D squat model and incorporated the effect of water entrapment,
lateral traction and elastic foundation on the stress intensity factor by using the
displacement at the crack tip, using an extended finite element modelling (X-FEM)
technique. In all of the analyses, the effect of loading eccentricity and head wear were
not considered.
The growth rate of multiple rolling contact fatigue cracks under the influence of tensile
longitudinal stresses in the rail was investigated by Fletcher et al [27], who found that
multiple cracks shield each other, thus reducing the crack growth rate. The crack growth
rate was found to increase as the bending stress increased. They studied the influence of
crack spacing for longer cracks using a boundary element method to investigate SIF (KI)
at the crack tip. It was found that KI is a function of crack spacing and that wider crack
spacing is directly proportional to KI, which leads to an increase in the cracking growth
of RCF cracks.
Dutton et al [21] presented a bending model to investigate the effect of an elastic
foundation on the stress intensity factor. The results suggested that for smaller cracks
vertical bending does not contribute significantly to crack growth, and that cracks larger
than a 30 mm radius are driven by vertical bending [19]. The modelling approach did
not consider crack closure and this resulted in negative values of KI.
Sandstrƶm et al [111, 112] investigated the risk of rail breaks from a mechanical and
statistical point of view. The influence of impact loads from flat wheels was particularly
considered. The flats increased the bending moments (and resulting stresses). However
depending on the impact position it did not have to influence the growth of an existing
crack. Hence, the influence on fatigue crack growth was minor. In contrast, bending
stress was a major concern in regards to the final fracture.
In an important study, Zerbst et al [124] conducted a damage tolerance investigation of
the rail for a typical squat defect approximating a semi-elliptical shape with different
sizes and an orientation of 70o on a rail running surface with a head wear of 10 mm. The
effects of the contact patch offset due to hunting, and the wear of wheel / rail profiles
were considered. The out-of-plane tearing mode was not investigated. Furthermore,
Chapter 2. Literature review
53
bending of the rail head-on-web due to the eccentricity of the contact load was not
considered in the crack growth investigations conducted by Fletcher et al [27, 119] and
Dang Van et al [28]. However, Jeong et al [6] included this behaviour in an
examination of the growth behaviour of reverse detail fractures that were initiated at the
lower gauge corner of heavily-worn rail. Kapoor et al [23] stated that relatively little
work had been done on large cracks that approach a critical crack length and may
rapidly propagate and result in rail failure. This effect is of increased importance under
high axle load conditions and for increased head wear [1].
More recent studies by Ranjha et al [3, 4] reported the occurrence of tensile stresses
resulting from lateral bending of the rail head and a localised response of the rail head-
on-web bending directly under wheel loading, resulting in tensile bending stresses at the
underhead radius. This effect is highly localized and is additional to the stresses
generated due to bending of the whole rail profile (so called global bending). These
stresses can initiate a crack at the underhead radius. In addition to head wear the
magnitude and direction of the L/V ratio (lateral (L) to vertical (V) load), the effect of
the contact patch offset (CPO), the foundation stiffness, and the residual and thermal
stresses were considered. The magnitude of these stresses may increase with increasing
rail head wear. In practice, as a result of variations in wheel-rail contact conditions, the
peak tensile stress at the underhead radius in both the gauge and field side of rail may
fluctuate considerably. A typical tension spike at the rail underhead radius occurs in a
narrow band corresponding to approximately 100 mm of wheel travel. Increases in
wheel-rail contact offset from the rail centreline and lateral forces, i.e. the ratio of lateral
to vertical loads (defined as the L/V ratio), that occur as a result of vehicle steering,
curving and hunting, may result in an increase in the longitudinal tensile stress. In
addition, the depth below the running surface at which the stress becomes tensile
decreases, which means that the region subjected to tensile stresses extends further from
the rail underhead radius towards the rail surface, and could result in an increased
tendency for Mode I crack growth behaviour in any pre-existing rolling contact fatigue
cracks [3, 4].
Chapter 2. Literature review
54
2.7 Wear fatigue interaction
The āWhole Life Rail Modelā developed by Kapoor et al [18, 23] and Dutton et al [21]
describes the crack growth rate during the life of the RCF crack from initiation to final
fracture, as shown in Figures 1.4 (a) and (b) and was described in detail in introductory
chapter [20, 22, 23, 30, 31]. A rapid fracture resulting from a transverse defect (TD)
such as that shown in Figure 1.4b [22, 25, 26], may occur if a RCF crack turns down at
a certain depth from the rail head, additionally driven by the tensile bending stress. The
extensions of āWhole Life Rail Modelā application to practical situations are described
next.
Fletcher et al [118] reported that wear affects crack propagation. Figure 2.24 describes
the effect of wear, in which a thin layer of material removed by wear truncates the
crack. Crack growth is the difference between fatigue led crack propagation minus
crack mouth truncation due to wear, as was represented by the expression given below
[22, 23]. Further, wear will also change the contact geometry and thus influence both
wear rate and RCF damage magnitudes.
Crack growth = crack tip propagation ā crack mouth truncation
Figure 2.24 Effect of wear on crack growth [22, 23]
The modification to the Paris equation is
( )
.
/ (2.35)
Chapter 2. Literature review
55
High wear rate can lead to the removal of cracks. Grinding is a standard practice to
control rail degradation.
Franklin et al [78] presented the modeling of wear and crack initiation in rails. Different
wear rate levels were selected to analyze the effect of wear on crack growth rate, as
shown in Figure 2.25.
Figure 2.25 Life cycle of a crack in a rail [125]
If the wear rate is high, as per level 1, no crack formation is evident. At level 2, the
crack is truncated by wear and the crack will reduce in length or disappear. At level 3,
the crack growth rate is positive and the crack will initiate and propagate leading to rail
fracture [78].
Generally rail can fail by either wear (loss of profile) or by RCF (long surface cracks or
a complete rail break). The length of life affected by wear (line B) decreases with an
increase in the wear rate, whereas the RCF life (line A) increases as the net crack
growth rate drops. The superimposing of these two lives gives the replacement life, as
shown in Figure 2.26. The peak on the curve represents the point of maximum life.
Historical experience reveals that it is difficult to maintain operation at the point of
maximum life, as variable operating and loading conditions reduce the optimum life
Chapter 2. Literature review
56
point. Both A and B have the same life, but B is inherently safer. These rail life
management strategies are valuable for the railway industry in terms of both rail safety
and economic benefits [22, 23, 25].
Figure 2.26 Strategies for rail life management [22, 23, 25]
2.8 Allowable wear limits
Marich [38] measured rail stress levels under both laboratory and field conditions to
define acceptable rail wear limits for high axle load conditions, based on the fatigue
behaviour of the rail material. Marich found an acceptable head wear limit of 27 mm for
600-800 m radius curve at an L/V ratio of 0.3, deemed to be outward for rail material
with a fatigue strength of 240 MPa. The head wear limit could be reduced to 20 mm as
an additional tensile stress of 80 MPa was added, due to the differential thermal
stresses. It was found that the rail wear limits proposed by Marich could be considerably
greater than those currently accepted in practice [33]. Marich [36, 37] identified the
underhead radius (UHR) as one of the most critical locations in determining the head
wear limit in the presence of a heavily worn rail profile. This work did not take into
consideration the possible impact of RCF damage. Subsequent research by Mutton et al
A B
Chapter 2. Literature review
57
[1] demonstrated that the approach taken by Marich was therefore non-conservative in
the presence of RCF damage.
Jeong et al [126] undertook rail strength investigations in relation to rail wear limits
based on the fracture strength of the internal transverse defects, classified as detail
fractures in North American Railroad systems. This study shows that for safe operation
on railroad track, allowable head wear limits should be estimated on the basis of
fracture strength. The research work conducted by Jeong et al [126] predicted head wear
limits for the lightest rail section. The head wear limits were estimated as 1.27 cm head
height wear or 1.52 cm gauge face wear, based on an assumption that a rail is inspected
for internal defects every 20 million gross tonnes (MGT). The limitations of this
research, to be used for the estimation of wear limits, include the accumulated MGT, the
axle load for different rail sizes, materials and also the above rail parameters.
Lyon et al [48] reported on a method to estimate rail wear limits based on a fracture
mechanics approach. A limitation of this study was that contact stresses were not
considered, based on an assumption that compressive stresses do not contribute to crack
growth. Also the axial (longitudinal) component of Hertzian stresses was calculated to
be sufficiently low that it would be neglected at depths where detail fractures (DFs) are
initiated.
Ranjha et al [17] examined unstable crack growth for long cracks at the gauge corner of
rail resulting from multiple GCC. In that investigation, a large surface crack was
introduced at the gauge corner. Crack growth behaviour with different contact loads
applied to the rail head at varied locations was evaluated in respect to different rail worn
profiles. The influence of tensile bending stresses at the underhead radius, associated
with contact loads, along with a combination of the local response of the head-on-web
bending and the lateral bending of the whole rail profile [3, 4], were included. As the
crack grows, the growth rate is dependent on whether it is driven by bending stresses or
contact stresses. The bending stresses associated with this behaviour are significant in
phase 3. It was predicted that the presence of high tensile underhead radius stresses
could lead to catastrophic rail failure as a result of the rapid fracture of pre-existing RCF
cracks extending to the underhead radius (UHR) region.
Chapter 2. Literature review
58
Estimation of head wear limits is critical for rail performance and plays a vital role in
terms of maintenance activities throughout the life cycle of a rail. In practice, a head
wear limit of 68 kg/m rail was stated by Duvel et al [13] to be in the range of 20 to 15
mm in curves of decreasing radii. The current head wear limit is set at 22 mm for
tangent track and 15 mm for tight rail radius curves, as reported by heavy haul railways
in Australia [33]. Typically, head wear (HW) of 20 mm is allowed for Deutche Bahn
(DB), as reported by Zerbst et al [124]. Therefore, head wear limit estimation and
predictions of rail life based on wear rate are very important research topics requiring
careful consideration. Investigation into rail parameters, including the different section
curve radii and geometric features offer huge potential for future research.
2.9 Material grades in high axle load rail operations
Rail material grades play a significant role in performance in terms of wear and rolling
contact fatigue. Selecting a high strength rail grade can resist wear and increase the
service life of rails, but it may be uneconomic for some railway operations. For high
axle load rail operations, high strength, heat-treated rail grades such as a plain C-Mn
Head Hardened (HH) grade [80], a Low Alloy Heat Treated (LAHT) grade [127], and a
Hypereutectoid (HE) heat treated grade [81] are commonly used, as these provide
increased resistance to damage in the form of wear and rolling contact fatigue [2, 128-
130]. Results for hardness distributions along the vertical transverse at the rail centreline
are plotted for high strength rail steels, as shown in Figure 2.27. Hardness throughout
the head of these rail types decreases with an increase in depth below the outer surface.
For Eutectoid and 1% C HE rail grade, near surface hardness varies from 375-430HV.
These grades show a trend in reduction of hardness with depth, particularly in head
hardened rail [2].
The proof stress (yield stress) for Head Hardened rail grades is around 600 MPa,
increasing to 1100 MPa for the Hypereutectoid (HE) heat treated grade, as shown in
Figure 2.28. The relationship between proof stress and hardness for low alloy heat
treated (LAHT3) rail grade is also shown in Figure 2.28. Fatigue limits are calculated
Chapter 2. Literature review
59
through rotating bending tests. Figure 2.29 presents the relationship between the tensile
strength (fracture stress) and fatigue limit [130].
Figure 2.27 Hardness distribution along vertical transverse at rail centerline [2]
This shows that with an increase in tensile strength the fatigue limit increases. Although
the ductility and the surface roughness of the material are the contributing factors that
influence the fatigue resistance generally the Hypereutectoid, (HE) rails (0.9%- 1% C)
have high fatigue limit, therefore fatigue defects are less likely to originate. These
calculations are for material on the rail surface. With depth, the hardness decreases, as
does the tensile strength and, ultimately, the fatigue limit. Therefore a thorough analysis
of hardness testing should be conducted for rail grades under investigation for an
estimation of fatigue life.
Alwahdi et al [128, 129] conducted an investigation on the effect of rail material grades
(pearlitic and bainitic) and surface roughness on the wear of rail steels. The preliminary
results showed that the wear rate of pearlitic rail steel was found to be lower in
comparison to bainitic rail steel at the similar hardness level.
250
270
290
310
330
350
370
390
410
430
450
0 5 10 15 20 25 30 35 40 45 50
Har
dnes
s (H
V20
)
Distance from surface (mm)
HE3LAHT3HH (AS1085.1)
Chapter 2. Literature review
60
Figure 2.28 Relationship between yield strength and hardness in eutectoid and hypereutectoid heat treated rail steel [2]
Figure 2.29 Relationship between tensile strength and fatigue limit [130]
400
500
600
700
800
900
1000
1100
1200
250 300 350 400 450
0.2%
Pro
of st
ress
(M
Pa)
Hardness (HV)
HE3LAHT3HH (AS1085.1)
400
420
440
460
480
500
520
540
560
580
600
1000 1100 1200 1300 1400 1500 1600
Fatig
ue li
mit
(MPa
)
Tensile strength (MPa)
HE3 (1.0%C)HE (0.9%C)DHH (0.8%C)
Chapter 2. Literature review
61
Residual and thermal stresses 2.9.1
Generally, there are two methods of rail straightening which affect the residual stresses
during rail manufacturing, namely roller [80] and stretch straightening [131]. In
addition, repeated rolling contact between wheel and rail is another important parameter
that can lead to a complex distribution of residual stresses in the rail. Tensile residual
stress at the rail underhead region could be detrimental to fatigue behaviour. The
differential temperature due to seasonal effect is another factor that influences the stress
state at the UHR in continuously welded rail (CWR) due to buildup of thermal residual
stresses, which are uniformly distributed in the cross section.
2.10 Summary
This chapter reviewed the basic concepts of contact mechanics, in particular the Hertz
contact theory, Hertzian contact stresses, plastic flow, shakedown, wear and rolling
contact fatigue, from the point of view of their definitions and their relationship with
their application in wheel-rail contact mechanics. It provided definitions and
terminology for rail defects that are found widely and described a method of fatigue and
fracture analysis. It also briefly outlined theories related to crack initiation based on
fatigue analysis. Furthermore crack propagation and fracture analyses based on linear
elastic fracture mechanics were carefully reviewed. Separate sections were devoted to
the review of wear fatigue interaction, allowable wear limits and rail material grades
under high axle load conditions.
This page is left intentionally blank
Chapter 3. Finite element model development
63
Chapter 3
Finite element model development
3.1 Introduction
This chapter presents the basics of finite element modelling strategy, which was used to
study the failure mode, as described in chapter 1. The proposed numerical models are
based on the Finite Element Method (FEM), Multi-axial Fatigue Analysis into
ABAQUS through UVAR-M (User output variable) and an extended finite element
method (X-FEM).
Firstly, a model of a rail section was constructed using the Finite Element Method
(FEM), to calculate the stress state around the underhead radius reference point by
considering the loading applied to an elliptical Hertzian contact patch on a rail surface
with a fully slipping condition, to give a worst case.
Secondly, several new finite element (FE) models were developed to study the effect of
worn profiles. These will be explained in chapter 5. The FE models developed to
calculate stress state were extended to evaluate fatigue damage around the underhead
radius (UHR) reference point. The Dang Van criterion was programmed into
FORTRAN-code and implemented into ABAQUS through UVAR-M (User output
variable), in order to identify any potential fatigue damage. This model development
will be explained in detail in chapter 6.
Thirdly, a numerical model based on the extended finite element method (X-FEM), was
used to examine the unstable RCF crack growth behaviour of long cracks in rail. The
multiple cracks introduced at the gauge corner of the model in the transverse orientation
Chapter 3. Finite element model development
64
had the approximate shape of typical long and turned down head checks that occur
under heavy haul conditions (Figure 1.2). These modelling details will be explained in
chapters 7 and 8.
3.2 Finite element model development
Geometric model 3.2.1
The geometric model used a 68 kg/m rail section corresponding to AS 1085.1 [80], a
rail section commonly used in Australian heavy haul railways, as shown in Figure 3.1.
The rail height is 185.7 mm and the minimum thickness of the rail web is 17.5 mm. The
width of the rail head and the base of the rail are 74.6 mm and 152.4 mm respectively.
The vertical and lateral bending inertia for a typical unworn or new rail are 39.4 x 106
mm4 and 6.02 x 106 mm4 respectively as reported in AS1085.1 [80]
Figure 3.1 68 kg/m rail [80]
Chapter 3. Finite element model development
65
The rail is installed in track at the relevant cant (or inclination) as was shown in Figure
2.2. The rail inclination is 1:40. In the current model the rail inclination was not
considered as the Hertzian contact patch was modelled instead of the actual wheel. It is
customary to do FE analysis relative to rail axis.
Flow diagram of the FE model 3.2.2
The detailed schematic for FE model development based on rail geometry, material
properties, loading and boundary conditions is shown in Figure 3.2.
Figure 3.2 Flow diagram for FE model solution technique development based on rail geometry, material properties, loading and boundary conditions
-nodes 134399
Chapter 3. Finite element model development
66
Track modelling 3.2.3
The IHHA (International Heavy Haul Association) publication [132] provides more up-
to-date guidelines on track designs for heavy haul operations. Types indicated in the
Australian Standards for the heavy haul systems for which the current work is
applicable suggest concrete sleepers at 600 mm spacing, elastic fasteners (resilient
fastenings generally Pandrol "e-clip"), resilient high density polyethylene (HDPE)
rail/sleeper pads, 68 kg/m continuously-welded rail, an axle load of 35 tonnes and
ballast with a nominal depth of 300 mm [80, 89, 132, 133]. Ballast depths are typically
300 mm at track construction but sometimes get deeper, due to the lifting/tamping
methods that are used for line and top re-alignment. However, current ballast depths and
the condition of the ballast will vary as a result of track degradation and ongoing track
maintenance activities over the lifetime of the infrastructure. For this reason, one aspect
of interest in this thesis has been to investigate the effect of ballast pumping and
additional effects such as degradation and looseness in sleepers, fasteners, rail pads and
subsequent variations in track stiffness levels on the response of the rail. This was
achieved by applying varying stiffness values to the elastic foundation in the lateral and
vertical directions of the rail, as will be described in the next section on FE modelling.
Finite element model 3.2.4
A finite element (FE) model for this analysis was developed using the commercial finite
element package ABAQUS 6.11-2. Only one rail was modelled to reduce computation
time. An elastic stress analysis was used to calculate the stress state around the
reference point by considering the contact loading applied at the midpoint of the rail
span (in longitudinal direction at x = 2100 mm) in between two middle sleepers at a
contact patch offset as shown in Figure 3.3 (a) and (b).
Elastic foundations were used to simulate the effect of rail pad, sleepers and ballast.
These were acting directly under the foot (wide bottom portion) of the rail and on both
sides of the rail foot area at the location of the sleepers, which were spaced at 600 mm
intervals along the length of the rail, as shown in Figures 3.3(a) and (b). In the vertical
Chapter 3. Finite element model development
67
direction, a stiffness / area (Kv) was applied to a contact area of 230 mm x 152.4 mm
underneath the rail, while in the lateral direction the stiffness / area (KL) was applied to
an area of 230 mm x 7 mm on both sides of the rail foot. Some key model parameter
values are given in Table 3.1. The values of all the parameters used in the model were
based on typical track designs used in Australian heavy haul rail systems.
Figure 3.3 The model description: (a) Top view of single rail supported with discrete elastic foundations; (b) Cross-section view of the rail, the measurement location and the elastic foundations
(b)
(a)
S
Chapter 3. Finite element model development
68
Table 3.1 Model parameter values
Model Parameters Values
Rail span, S (mm) 4200
Support length, Ls (mm) 230
Support width, Ws (mm) 152.4
Axle load (kN) 343.4
Contact patch offset (mm) 0, 15, 30
L/V ratio 0, 0.2, 0.4
Thermal expansion coefficient, Ī± (/oC) 1.2e-5
Youngās modulus, E (MPa) 209,000
Poissonās ratio, 0.3
Boundary conditions 3.2.5
The rail was constrained against the longitudinal translation (u1 =0) and rotation (uR1 =
0) at both ends. The vertical and lateral translations were constrained by the vertical and
lateral stiffness applied at the sleeper locations. For an S = 4200 mm span of rail, as
shown in Figure 3.3(a), the resultant stresses considered at the centre location of the
cross-section were almost unchanged as a result of boundary condition changes at both
ends of the rail.
Loading 3.2.6
The wheel load was applied on the rail assuming a fully slipping Hertzian contact patch.
The wheel vertical load (V) was applied through a Hertzian pressure distribution over
an elliptical contact area of 125.6 mm2 (semi major axis, a = 10 mm, semi minor axis, b
= 4 mm as was reported by Mutton et al [1]). The pressure distribution is for heavy haul
operations in Auatralia for an axle load of 35 tones; the current analysis has used the
same axle load as mentioned in Table 3.1. The Hertzian pressure distribution is given by
the following expression:
Chapter 3. Finite element model development
69
( ) ā.
/ (3.1)
(3.2)
Vehicle curving and hunting 3.2.7
Vehicle curving and hunting behaviour was of interest in this doctoral study due to its
effect on contact patch position and lateral forces. This is exacerbated if contact moves
away from the centre of the rail head cross-section. In real life the contact patch size and
shape vary as the contact moves towards the gauge corner. Figure 3.4a shows a
simplified representation of the contact patch shape versus different offsets (0, 15, and
30 mm). Depending on the curvature, double contacts are possible as well as significant
departures from the ideal Hertz assumption.
The focus of this study was the stress state of the underhead radius, as a result of
localized section bending behaviour (lateral bending of whole rail profile and localized
vertical and lateral bending of head-on-web) on sleepers rather than the contact stresses
close to the running surface. Therefore, to minimize the effects of the contact patch size,
it was kept unchanged at the different offsets on the rail running surface, as shown in
Figure 3.4b. Likewise, contact pressure distribution varies as the contact patch moves
towards the gauge corner. In order to minimize this effect, ellipsoidal Hertzian contact
pressure distribution was used at different contact patch offsets. However the effect of
changes in contact patch size and shape are only small. The underhead radius stresses
showed a maximum difference of 5%, as a result of 20% variations in the values of
major and minor semi axes due to subsequent changes in contact geometry and
maximum contact pressure.
Salehi et al [43, 45] have also made such a simplification in contact patch offset and
contact pressure distribution. They reported that the underhead radius position is far
away from the contact patch and that the difference in stress distribution at the
underhead radius between uniform and ellipsoidal contact pressure is less than 1%. The
Chapter 3. Finite element model development
70
underhead radius stresses are associated with a combination of lateral head bending of
the whole rail profile and localised bending of the head-on-web on the elastic
foundation due to contact patch offset. Mutton et al [1] and Marich [36-38] also
reported the contact patch offset from the centre of rail head cross-section with similar
loading conditions to those stated above. A maximum contact patch offset of 30 mm
was considered in this thesis, as shown in Figure 3.5. The contact loads were located at
the mid-point of the rail span (in longitudinal direction at x = 2100 mm) in between the
two middle sleepers (Figure 3.3a), at different contact patch offsets. The major semi
axis, a, was in the longitudinal direction and the minor semi axis, b, was in the lateral
direction at the midpoint of the rail span (x = 2100 mm), as shown in Figure 3.5.
Figure 3.4 Variations of contact patch sizes and shapes
In real life
Current analysis
Same size and shape
(a)
(b)
Chapter 3. Finite element model development
71
Figure 3.5 Representation of a contact patch offset (CPO) from the center of rail head cross-section for modelling eccentric loading
Vehicle dynamics studies [64, 134] have described the direction of the lateral shear
traction, which can be either inward or outward depending upon the curving, steering
and hunting mechanism of the vehicle. The high rail will experience outward lateral
shear traction (towards the field side) from the leading wheelset during curving but, due
to the rigid bogie side frame, the trailing wheelset will try to readjust itself and produce
an inward lateral shear traction (towards the gauge side) on the rail head surface, as
shown in Figure 3.6. On a tangent track, the hunting mechanism due to vehicle
dynamics is considered to be the main source of lateral shear traction.
The lateral forces are influenced by many factors, such as wheel-rail profiles, curve
radius, lubrication conditions and vehicle speed. The calculation of these forces was
beyond the scope of this thesis, and the simulations have been conducted for several
L/V ratios to cover all possible cases. The Hertzian pressure distribution for local
coordinate systems (XYZ) can be found in Figure 3.7(a) and (b). The contact pressure
PZ (X, Y) had contributions from both the wheel vertical load (V) and the lateral load
(L) to the contact patch and was applied through a Hertzian pressure distribution over an
Chapter 3. Finite element model development
72
elliptical contact area with an angle in the vertical direction, as shown in Figure 3.7
(a) and (b). The normal and tangential forces (FZ and FY) at the contact point in the YZ
plane for the local coordinate system (XYZ) can be expressed as
FZ = Vcos ā Lsin (3.3)
FY = Lcos + Vsin (3.4)
Figure 3.6 Representation of inward and outward lateral shear tractions by leading and trailing wheelsets respectively during curving on high rail of a 600 m radius curve (adopted from Xiao et al [134])
The lateral load (L) could be caused by both the steering effects on the curved track and
the hunting of the vehicle due to vehicle dynamics, and normally these are not present at
the same time. This can produce lateral shear traction in the contact area, in combination
with the elevated friction levels due to the absence of any lubrication at the wheel-rail
interface. The magnitude of the loads can be additionally impacted by the vehicle
dynamics [60, 64]. The studies conducted by Salehi et al [43, 45] revealed that the
longitudinal traction does not affect the stress state at the underhand radius, so it can be
eliminated. In order to perform the sensitivity analysis on the effect of lateral traction,
the ratio of lateral load to the vertical load (L/V ratio) can be varied, defined as the
lateral traction coefficient. Accordingly, the contact patch offset and the L/V ratio (ratio
Chapter 3. Finite element model development
73
of lateral to vertical load) are representative of load cases that approximate the hunting
and steering behaviour of wagon. Cases with an L/V ratio of (0, 0.2 and 0.4) and contact
patch offsets (0, 15, 20 and 30 mm) were parametrically simulated.
Figure 3.7 The rail-wheel contact loads in detail (a) wheel loads [the picture to the left is the original work]; (b) Hertzian pressure distribution at the contact patch [picture to the right is adopted from Ekberg et al [62]]
Mesh development 3.2.8
The rail 3D FE model with dense mesh at the contact patch is shown in Figure 3.8. The
model used 8 node 3D solid elements with reduced integration (C3D8R). The dense
mesh at the contact patch used 4 node 3D tetrahedral solid elements (C3D4) [135]. In
order to reduce the element numbers, the mesh refinement was updated to obtain a fine
mesh at the rail head around the contact patch area only, at all the three offset locations
and at the underhead radius location where the stresses were being calculated. Coarse
mesh was used along the length of the rail and on the web and foot areas. The model has
177,074 elements and 135,399 nodes. The maximum element size was 20 mm in the
rail, but the smallest element size was 0.007 mm in the contact patch region.
(a) (b)
Chapter 3. Finite element model development
74
Figure 3.8 Mesh development with dense mesh at the contact patch
3.3 Model Validation
Field data 3.3.1
Field measurements for the model validation were based on track instrumentation
conducted by Mutton et al [1, 33] and Bartle et al [32], who measured longitudinal
stress at the underhead radius on both the field and gauge side, as well as vertical
stresses in the rail web and longitudinal stresses at the top of the rail foot at several track
locations subject to high axle load rail traffic. Vertical wheel loads were also measured
using conventional shear bridges. For this activity, the detailed positions of the strain
gauges are shown in Figure 3.9 (a) and (b). These were applied to the relevant position
on the welds (Figure 3.9b), and on the corresponding position on the rail (Figure 3.9a)
at a distance of approximately 1200 mm from the instrumented welds, in the direction
facing the approaching loaded traffic.
The field data used for the model validation are shown in Figure 3.10. Longitudinal
stress in the rail head was measured at 8 locations (4 field / 4 gauge), as given in
Chapter 3. Finite element model development
75
Appendix B. A typical pattern shows rail going into compression due to the normal
bending action, with a tensile spike associated with lateral bending of the whole rail
profile and localised vertical and lateral bending of the head-on-web, as shown in Figure
3.10. The variations in tension peaks for each wheel can be attributed to differences in
combinations of contact patch positions, vertical wheel loads and the lateral traction
forces for individual wheels. The magnitude of the spike is also dependent upon rail
wear conditions. It was found that the tensile bending stress at the gauge side underhead
radius of the rail reached a peak value of about 100 MPa when the wheel was directly
above the strain gauge position (underhead offset of measurement point 20 mm). This is
a 609 m radius curve as mentioned in the caption and the speed and resulting cant
deficiency is unknown. The position of the offset measurement point has been defined
in Figure 3.3 and the in-situ position is shown in Figure 3.9
Figure 3.9 The strain gauges at the measured points [33]
(a)
(b)
Chapter 3. Finite element model development
76
Figure 3.10 Measured longitudinal stress response under in service loading; example shown for high rail in 609 m radius curve [1, 32]
The corresponding stresses on the field side underhead radius of rail reached a peak
value of about 83 MPa. For different wheels of a single train the magnitude of peak
tensile stress varied considerably at the both gauge and field side underhead radius
positions of the rail as given in Appendix B.
Hertzian contact stresses 3.3.2
The FE model was validated by comparing the stress field below the centre of a circular
Hertz contact patch with those obtained using an analytical solution. A circular Hertzian
contact patch with a radius of 6.8 mm, supporting a vertical load of 171.7 kN, was used
in the FEA model and also for the calculation of the analytical part [58].
Tension spikes
Compressive stresses
Time (sec)
Time (sec)
Chapter 3. Finite element model development
77
Figure 3.11 Stress distribution in the rail head versus depth caused by Hertz pressure acting on a circular area (a=6.8 mm, Fz =171.7 KN)
The FEA results revealed that the longitudinal stress at the centre of the contact surface
had a maximum difference of 4% compared to the analytical solution, as shown in
Figure 3.11.
Octahedral shear stresses 3.3.3
The octahedral shear stress ( ) was evaluated analytically as:
ā
,( ) ( ) ( ) - (3.5)
The octahedral shear stress distribution along the centreline below the contact area
showed a maximum difference of 5% between the analytical solution [58] and the
-3
-2.5
-2
-1.5
-1
-0.5
0-1.5-1-0.50
Dim
ensi
onle
ss d
epth
(z/a
)
Ratio of stresses to Pm (MPa)
Ļz/Pm FEMĻz/Pm ANALYTICĻr/Pm FEMĻr/Pm ANALYTICĻ/Pm FEMĻ/Pm ANALYTICĻĪø/Pm FEM
Chapter 3. Finite element model development
78
current FEA prediction, as shown in Figure 3.12. This variation is within an acceptable
range when compared to the FEA solutions as reported by [1] and [69].
Figure 3.12 Octahedral shear stress ( ) of a rail subjected to a non-uniform (Hertzian) contact pressure (a = 6.8mm, Fz = 171.7 KN)
Effect of elastic foundation 3.3.4
Appropriate values of stiffness / area of elastic foundation in the vertical and lateral
directions were chosen using evaluations of the deformation of a rail head in both the
vertical and lateral directions against field measurements. Table 3.2 shows some of the
compared results. The finite element analysis (FEA) results indicate that, with the stiffness /
area of (KV) = 1.0 N/mm3 and (KL) = 1,740 kN/mm3, the vertical deformation of the rail
head is about 2-3 mm and the lateral deformation is in the order of 1 mm (for Fz = 171 kN).
These deformation values were as expected previously in field or laboratory measurements
-16
-14
-12
-10
-8
-6
-4
-2
00 100 200 300 400 500 600
Dep
th b
elow
rai
l sur
face
(mm
)
Octahedral shear stress (MPa)
FEM
ANALYTIC
Chapter 3. Finite element model development
79
[1]. For simplicity, stiffness was taken as uniform and was assumed to be linear, even
though, in reality it is non-linear due to different behaviour in tension and compression. KV
was applied over an area of 35,052 mm2 (230 mm x 152.4 mm) and KL over an area of
1,610 mm2 (7 mm x 230 mm). With this, the vertical stiffness applied is 35 kN/mm and the
lateral stiffness applied is 2800 MN/mm.
Table 3.2 Results of simulation of rail deformation with respect to the support of elastic foundation
1. Elastic foundations: N/A. Boundary conditions: fixed at base and both sides of the rail foot at the location of sleepers. 2. Elastic foundations: Kv = 1 N/mm3. Boundary conditions: fixed in lateral direction at both sides of rail foot at the location of sleepers. 3. Elastic foundations: Kv = 1 N/mm3, and KL = 1740 kN/mm3. Boundary conditions: N/A. 4. Elastic foundations: Kv = 0.1 N/mm3, and KL = 17.4 kN/mm3. Boundary conditions: N/A. 5. Elastic foundations: Kv = 1 N/mm3, and KL = 1740 kN/mm3. Boundary conditions: N/A.
The lateral stiffness is applied to a very small area at both sides of the rail foot, as
shown in Figure 3.3b, to model the rail fastening applied on these locations. Therefore,
the stiffness values are much higher compared to the vertical stiffness value. Similarly,
the difference in deformation is a factor of about -3, whereas the difference in stiffness
is a factor of about 2000 (or 10000 if forces are considered) as shown in Table 3.2.
Chapter 3. Finite element model development
80
These boundary conditions, especially the vertical and lateral stiffness, provided some
constraint to the rail regarding rotation in the lateral and vertical directions.
Sensitivity analysis 3.3.5
A sensitivity analysis was performed to investigate the bending and deformation
behaviour using different support and boundary conditions, as shown in Table 3.2. Field
measurements [1] showed the lateral deformation to be of the order of 1 mm and the
vertical deformation to be of the order of 2-3 mm. Case 3 in Table 3.2 with L/V = 0.2
and 0.4 provides a sufficiently good match for both and these support conditions were
used throughout the doctoral investigation. The longitudinal stresses at the underhead
offset of measurement point 20 mm are plotted in Figure 3.13, and a tension spike can
be seen to occur directly under the wheel contact position. This behaviour has been
observed in field measurements, as shown in Figure 3.10.
Figure 3.13 Longitudinal stress at the underhead offset of measurement point 20 mm
The FE results were compared with in-situ measured stress data for loaded rail traffic
with a nominal axle load of 35 tonnes. The in-situ measured data showed variability in
-60
-40
-20
0
20
40
60
80
100
120
1800 1900 2000 2100 2200 2300 2400 2500
Lon
gitu
dina
l Str
ess (
MPa
)
Longitudinal Position (mm)
L/V=0 contact patch 15 mm offsetL/V=0.2 contact patch 15 mm offset L/V=0.4 contact patch 15 mm offsetL/V=0.2 contact patch 30 mm offsetField measurements
Chapter 3. Finite element model development
81
peak tensile stress values at the measurement location, which could be attributed to
variability in either one or a combination of the contact positions for individual wheels,
vertical wheel loads and the lateral traction forces that arose from the steering effects.
The variations could be due to wagon loading variations. The values of these parameters
were not determined during the in-situ measurements, and hence the current analysis
made use of peak stress data for a sample of 21 wheel passes, which represented a
portion of the total data set. The peak tensile stress at the measured location (the
underhead offset of measurement point 20 mm) from the in-situ measured data showed
100.6 MPa, which was a smaller difference of 6.8 % in the case of an L/V ratio of 0.2
and a contact patch offset of 30 mm. The other peak tensile stress values at the in-situ
measured location are also plotted against FE results and are showing sufficiently good
correlation, see Figure 3.13.
3.4 Thermal stresses
The differential temperature due to seasonal effect is another factor that influences the
stress state at the UHR in continuously welded rail (CWR). An approach to calculating
the thermal stresses due to variations of service temperature from stress free or neutral
temperatures was stated as
(3.6)
Where T is the deviation from the stress free temperature, E is the Youngās modulus
and is the thermal expansion coefficient, as given in Table 3.1. The thermal stresses in
the longitudinal direction of the rail due to either a cold or warm temperature changes
were included in the simulation using the commercial FE package ABAQUS 6.11-2.
3.5 Residual stresses distribution
The residual stress (RS) induced by the manufacturing process and repeated rolling
contact between wheel and rail is another important parameter that can change the stress
state in rails and is often included in the specification of the rail by the manufacturer
Chapter 3. Finite element model development
82
[81, 127]: hence the interest in examining the influence of residual stress distribution on
the development of RCF. The Australian Standard (AS1085.1) [80] suggests a typical
distribution of longitudinal residual stresses in roller-straightened rail, which presents at
the underhead region with changes from -65 to 25 MPa.
Figure 3.14 Rail section for residual stress distribution
Worn profile
Original profile (not to scale)
Chapter 3. Finite element model development
83
Table 3.3 Residual stress scatter in the worn rail section given above [136]
Residual stresses, (MPa)
Rail section Ļx Ļy Ļz Ļxy Ļxz Ļyz
1 -250 -300 -300 0 0 -40
2 -150 -100 -200 0 0 -80
3 -150 -100 -100 0 0 0
4 -50 -200 -100 0 0 0
5 -87.5 -87.5 -87.5 0 0 0
6 -162.5 -80 -80 0 0 0
7 -12.5 -12.5 -12.5 0 0 0
8 100 0 0 0 0 0
For this thesis, the residual stress distribution was based on the measurements
conducted by Magiera [136]. As rail geometry is quite complex, the residual stresses
were considered in terms of the six stress components for each of the regions shown in
Figure 3.14. The input residual stresses are summarized in Table 3.3. The rail grades
used in the field are both roller and stretch straightened, therefore RS values used may
not be representative of actual rail grades operated in the field. They were applied by
defining them as initial conditions in an input file created in the FE code ABAQUS
6.11-2. For the first step in the rail FE model calculation, residual stresses became the
initial conditions and no contact loads were applied for this step, in order to prevent any
contraction and bending.
3.6 Summary
In this chapter, the author has presented the methodology and validation of the finite
element model, which was used to study the failure mode as described in chapter 1. The
proposed numerical model is based on the Finite element method and is validated with
in-track field data. The effect of the contact patch offset and L/V ratio, representative of
Chapter 3. Finite element model development
84
curving and hunting, was included and directly correlates with a tension spike due to
localised bending behaviour of rail. In the current model, allowance was made for the
effect of foundation stiffness, thermal stresses arising from variations around the
nominal stress-free or neutral temperature and the residual stresses, in order to
determine the applicable stress limits.
Chapter 4. Underhead radius stresses
85
Chapter 4
Underhead radius stresses
4.1 Introduction
The work presented in this chapter describes the first step in modeling the effect of local
rail bending stresses. The development of a finite element model of the rail, as described
in chapter 3, is applied to demonstrate the rail local bending behaviour at the underhead
radius, influenced by contact, bending and seasonally dependent thermal stresses. The
finite element model validation in chapter 3 demonstrated that the longitudinal stresses
in the rail underhead radius position are of special interest, as tension spikes have been
identified at this location during in-track measurements under high axle load conditions
[1]. The analysis revealed that the magnitude of the tension spike was highly dependent
on several service conditions: the contact patch offset from the rail centreline, the ratio
of lateral (L) to vertical (V) loads, the direction of lateral shear traction, foundation
stiffness and seasonal temperature variations.
In addition to the vertical bending of the whole rail cross-section on the elastic
foundation, the rail also undergoes lateral bending as well as vertical and lateral bending
of the head-on-web that was reported by Jeong [39], Jeong et al [40] and Orringer et al
[41]. The tension spike is a result of this effect, and corresponds to an additional local
bending stress due to the lateral bending of the whole rail profile and the localised
bending of the rail head-on-web. This effect is highly localised and is additional to the
stress generated due to vertical bending of the whole rail profile (the so called global
bending). Orringer et al [47] analysed the longitudinal bending stresses in the rail head.
They found the stress state to consist of five possible components: (1) vertical bending,
on-web bending, as was discussed in section 1.1.2, chapter 1. The field measurement
from Orringer et al [47] showed that, for tangent track, the longitudinal stress caused by
rail head-on-web bending is much smaller than that caused by rail vertical bending. The
warping term of the longitudinal stress is similarly small. On curved track, the
additional components of lateral bending and lateral head-on-web bending significantly
increase the longitudinal stress.
Eisenmann [34] measured the tensile stresses at the underhead location (at the lower
gauge corner) under field conditions within the German rail system, and also calculated
the theoretical tensile stresses at the gauge and field side underhead location for
eccentric and inclined loading. The local tensile stresses on the field side underhead
location under laboratory conditions and in the field due to lateral wheel loading were
measured by Sugiyama et al [35]. Marich [37] presented finite element analysis results
that demonstrated that the longitudinal stresses at the underhead radius are increased
when the contact patch moves away from the centre of the rail head cross-section on a
heavily worn rail. The increased stress level leads to fatigue failure and is a potential
contributor to rail defects. It was suggested that rail life could be improved if the contact
patch offset is controlled by adopting appropriate rail grinding strategies. Sugino et al
and Farris et al [93, 94] also reported the presence of large tensile stresses at the
subsurface (about 3-15 mm below the running surface inside of the rail head) of the rail
head. These stresses can become high and produce reverse detail fractures (i.e.
transverse defects which initiate at the lower corner of the gauge face of heavily worn
high rail), as was shown in Figure 1.7-1.12. It would be of interest to compare this
earlier work with what has been observed under high axle load conditions.
Jeong et al [6, 7] investigated reverse detail fractures by providing a close form solution
to calculate the longitudinal bending stresses, using simplified wear geometry on the top
and side of a rail head. A sensitivity analysis was undertaken to examine the influence
of rail head wear, residual stresses, thermal stresses and track stiffness on crack growth
rates. The scope of this research, which estimated crack growth rates for reverse detail
factures, included: the accumulated MGT, axle load for simplified wear geometry for
different rail sizes, materials and above rail parameters. Jeong [8] specifically
mentioned flow lips as an initiator, in addition to a possible contribution by higher
Chapter 4. Underhead radius stresses
87
longitudinal bending stresses at the lower gauge corner of the rail. The reverse detail
fracture failure mode is influenced by longitudinal bending stresses at the underhead
radius, as reported by Jeong et al [6].
Failures that initiate at the underhead radius of aluminothermic welds have been found
in Australia under high axle load conditions. Salehi et al [42-45] conducted a study
using multi-axial fatigue analysis to examine fatigue crack initiation in the vicinity of
the weld collar of aluminothermic welds. The results obtained permitted systematic
investigation of factors influencing fatigue occurrence based on careful FEM elastic
calculations, including, but not limited to, the contact zzonal local bending stress
superimposed due to the bending of the rail head on the web. The localisation of this
effect in a narrow band of 100 mm corresponded to the wavelength of this bending as
mentioned by Salehi et al [42-45]. This tension spike is very important in explaining
crack initiation at the underhead radius. The effect of head wear of weld profile and
inward lateral shear traction, which were not considered in these investigations,
exacerbate the failure at the gauge side underhead radius, and also in the area where the
tension spike generated shifts upwards from the underhead radius to the rail head
surface. This presents a potential risk for existing RCF cracks to turn down and become
transverse defects.
Mutton et al [1] modelled the elastic foundation effect in a 68kg/m worn rail section
laid on 8 discrete sleepers. Their model is capable of showing underhead radius stresses
due to lateral bending of the head and vertical and lateral bending of the head-on-web.
This effect is additional to the bending of a whole rail profile (so called global bending).
The boundary conditions and contact patch lateral movement created high longitudinal
bending stresses in the form of short duration tensile stresses at the underhead radius. It
was concluded that the tensile stress at the underhead radius increased with an increase
in head wear and L/V ratio, but the depth at which stresses became tensile reduced. This
high tensile stress on the gauge corner of rail head could cause a RCF initiated crack to
turn down to form a transverse defect. The underhead radius stresses were qualitatively
correlated with field measurements. For this activity, strain gauges were applied to the
relevant position on the welds, and on the corresponding position on the rail at a
Chapter 4. Underhead radius stresses
88
distance of approximately 1200 mm away from the instrumented welds, in the direction
facing the approaching loaded traffic.
Measurements conducted by Mutton et al [1] under high axle load conditions showed
that the tensile bending stress at the underhead radius occurs as a short duration peak
when the wheel is directly above the measurement position. In the field measurements,
it was reported that the tension spikes at the gauge side underhead radius reached a peak
value of about 100 MPa when the wheel was directly over the strain gauge position
(underhead offset of measurement point 20 mm). The corresponding stress on the field
side of the rail was about 83MPa, being 17% less than that in the gauge side of the rail.
With different wheels on a single train, the magnitude of peak tensile stress varied
considerably at both the gauge and field side underhead radius positions of the rail.
Greisen [53] and Greisen et al [137] presented an estimation of rail bending stresses
from real time vertical track deflection measurements, based on studies conducted by
Lu [138]. The long range (thermal, residual and bending) stresses, especially the cyclic
axial (longitudinal) stresses, usually dominated rail stresses and lead to fatigue crack
propagation and fatigue failure. Greisen et al [137] also reported that the longitudinal
bending stresses due to vertical bending could be a potential contributor to fatigue
failure in rail. The study focused on vertical bending, and the effects of lateral bending
and localised bending of the head-on-web were not considered.
In this chapter the longitudinal bending stresses as a result of additional lateral head
bending and localised vertical and lateral bending of the head on the web are being
taken into account. Vehicle dynamics studies [134] show that some wheels develop
inward shear traction (towards the gauge side) and some outward shear traction
(towards the field side) on the rail surfaces, which lead to differences in the tracking
behaviour of individual wheelsets during curving. Both these cases were considered for
this thesis. The effect of foundation stiffness and seasonal temperature variations were
also investigated. The results generated are described and discussed next. The work
presented in this chapter is a contribution to the existing literature [3]. The qualitative
effect of these underhead radius stresses on the propagation of pre-existing RCF cracks
and fatigue crack initiation at the underhead radius position will also be discussed. For
Chapter 4. Underhead radius stresses
89
this chapter, the longitudinal underhead radius stresses due to local bending were
calculated to form input to fatigue crack initiation and rapid fracture models, and the
development of these is detailed in chapters 6, 7 and 8.
4.2 Finite element model
The finite element (FE) model of nominal rail profiles was developed using the
commercial finite elements code ABAQUS (6.11-2) and validated with field
measurements. Only one rail was modelled to reduce computation time. An elastic
stress analysis was conducted for a stationary contact loading. The finite element
method was used to calculate the stress state around the reference point. The loading
was considered to be a fully slipping Hertzian contact pressure modeled on an elliptical
contact patch to give a worst case. The case of a wheel pass over a sleeper and then
between two sleepers was simulated and it was found that the case when the wheel is in
the middle of two sleepers gives the worst case. A detailed description regarding model
development and calibration was already provided in chapter 3. It should also be noted
that the details on input parameters such as loading, boundary conditions and support
characteristics will remain the same as was discussed in chapter 3 and are not repeated
here.
4.3 Effect of different contact patch offset and lateral tractions
The position of contact patch and the wheel rail forces depend on several parameters
and are not within the scope of this study. However to make these results useful, all
cases have been examined. A vehicle dynamics study would be required to determine
L/V ratios, contact patch offset and direction of lateral loading. The outputs of vehicle
dynamics studies would next be used to read the appropriate stress values.
The simulations were primarily conducted for tangent track with free rolling conditions
i.e. (L/V =0) on the centre of the rail head cross-section. In the case of a pure vertical
load (L/V = 0) on the centre of a rail head cross-section, there was no tensile stress
generated at the underhead radius in the gauge side of the rail, as shown in Figures 4.1
Chapter 4. Underhead radius stresses
90
(a) and 4.2 (a). The tensile stress at the underhead position showed an increase with
movement of the contact location to the gauge corner of the rail, as shown in Figure 4.1
(b). Figure.4.2 (a) shows that the peak tensile stress was 60 MPa at the underhead offset
of measurement point 34 mm for the contact patch offset of 15 mm with L/V = 0 (wheel
load of 172 kN). The vertical load eccentricity imposed local bending and additional
torsion of the rail head. The tensile stress at the underhead radius was attributed to the
localized bending behaviour of the rail head. The narrow band (around 100 mm) of this
effect corresponds to the wave length of this bending. When the eccentric position of
loading was increased from the contact patch 15 mm offset to 30 mm, the tensile stress
rose from 60 MPa to 140 MPa consistently.
Figure 4.1 Longitudinal stress (S11) contour; (a) centric loading; (b) eccentric loading on gauge side (contact patch 30 mm offset); (c) eccentric loading on gauge side (contact patch offset of 30 mm with L/V=0.4 towards gauge side)
(a) L/V=0, CPO=0 mm
(b) L/V=0, CPO=30 mm (c) L/V=0.4, CPO=30 mm
y z x
(UHR)
Chapter 4. Underhead radius stresses
91
Figure 4.2 Longitudinal stress at the underhead offset of measurement point 34 mm; (a) at different contact patch offsets (0, 15 and 30 mm) with L/V=0; (b) at contact patch 30 mm offset with different L/V= (0, 0.2 and 0.4)
When considering the hunting and steering behaviour of the rail, assuming the L/V =
0.2 and 0.4 with the contact patch 30 mm offset, the tensile stress takes place at the
gauge side underhead radius of the rail and increases consistently with an increase of
L/V ratios. As a result of the eccentric loads from L and V, the high tensile stress
imposed by the vertical and lateral bending of the head-on-web can clearly be observed
at the underhead radius, as shown in Figure 4.1 (c). The longitudinal stress at the
underhead offset of measurement point 34 mm for contact patch 30 mm offset with
-60
0
60
120
180
240
300
360
1800 1900 2000 2100 2200 2300 2400 2500
Long
itudi
nal S
tres
s (M
Pa)
Longitudinal Position (mm)
L/V=0L/V=0.2L/V=0.4
-60
-40
-20
0
20
40
60
80
100
120
140
1800 1900 2000 2100 2200 2300 2400 2500
Long
itudi
nal S
tres
s (M
Pa)
Longitudinal Position (mm)
contact patch 0 mm offsetcontact patch 15 mm offsetcontact patch 30 mm offset
(a)
(b)
34mm
Chapter 4. Underhead radius stresses
92
different loading ratios L/V of 0, 0.2 and 0.4, (wheel load of 172 kN) is shown in Fig.
4.2(b). The tension spike is about 324 MPa (for estimated torsional fatigue limits for
HH rail, t-1 = 205 MP) with a lateral load (L/V = 0.4), which is about two times higher
than without the lateral load (L/V = 0). It can be seen that the tension spike clearly
increases with an increase in the L/V ratio.
4.4 Stress state at the underhead radius and gauge corner
The behavior of the tension spike, in a range of positions at the underhead radius and
with depth in the rail head at gauge corner region, is considered next. Figures 4.2 a-b
show that the tension spike occurs in a narrow band (around 100 mm in length) and the
stress is highest at UHR surface. The variation of longitudinal stresses in this region is
of importance, as these high tensile stresses can cause cracks to develop and grow.
Figure 4.3 Longitudinal stress at different underhead offset of measurement points (20, 34 and 37 mm from longitudinal centerline of rail) at the contact patch 30 mm offset and L/V=0.4
Figure 4.3 shows the longitudinal stress distributions at the underhead offset of
measurement points 20, 34 and 37 mm from the longitudinal centreline of the rail. The
tension spike is highest with a peak value of 324 MPa when the underhead offset
measurement point is 34 mm, about 48% higher than that at the underhead offset of
-50
0
50
100
150
200
250
300
350
1800 1900 2000 2100 2200 2300 2400 2500
Lon
gitu
dina
l Str
ess (
MPa
)
Longitudinal Position (mm)
underhead offset of measurement point 37 mm
underhead offset of measurement point 34 mm
underhead offset of measurement point 20 mm
Offset
Chapter 4. Underhead radius stresses
93
measurement point of 20 mm. Beyond the underhead offset measurement point of 34
mm, the tension spike reduces slightly by 7% at the underhead offset measurement
point of 37 mm. Figure 4.4 shows the variation of longitudinal stress with depth on a
vertical plane at 20 mm offset from the vertical rail centreline towards the gauge corner.
Figure 4.4 Longitudinal stress distribution with depth on a vertical plane at 20 mm offset from the rail vertical centreline towards the gauge corner for the different contact patch offsets (0, 15, and 30) mm with L/V ratios (0, 0.4)
The contact patches are 0, 15 and 30 mm offset and the L/V ratios are 0 and 0.4. It can
be seen that, for L/V = 0.4, the longitudinal stresses are more tensile for the 30 mm
contact patch 30 mm offsetcontact patch 15 mm offsetcontact patch 0 mm offset
L/V = 0
Chapter 4. Underhead radius stresses
94
contact patch offset. They shift from compression to tension at a depth of 14 mm below
the contact surface, whereas for the contact patch 15 mm offset, the depth is about 20
mm. Where there is no lateral loading (L/V = 0), the tensile stress occurs at a depth of
38 and 34 mm for contact patches 15 and 30 mm offset respectively, showing small
sensitivity to changes in the contact patch position.
Figure 4.5 shows the longitudinal stress variation, with depths for different vertical
plane offsets of 16.5, 20, 25 and 35 mm from the rail vertical centreline towards the
gauge corner with L/V ratios of 0, 0.2 and 0.4 at the contact patch 30 mm offset. By
increasing the vertical plane offset from the vertical centreline of the rail toward the
gauge corner (L/V = 0.4), the longitudinal tensile stress increases, but the depth at
which the stress become tensile reduces from 20 mm to 2 mm. For the case with a
lateral loading of L/V = 0.2, the longitudinal tensile stress increases, and the depth at
which stress becomes tensile reduces from 29 mm to 5 mm. If there is no lateral
loading, i.e. L/V = 0, the longitudinal tensile stress increases, and the depth at which the
stress becomes tensile reduces from 38 mm to 14 mm. This behaviour can be verified
from the deformation of the rail, as shown in case 5 of Table 3.2. The lateral
deformation on the gauge side of the rail head at the contact patch 30 mm offset
increases 3.5 times as the L/V ratio changes from 0 to 0.4. The gauge side is in tension,
which will subsequently result in tensile stresses at the gauge corner.
The contours of the longitudinal stress on the cross-section of the rail head, with respect
to the contact patch offset of 30 mm with different L/V ratios, are shown in Figure 4.6.
The regions subjected to tensile stresses in the rail head are clearly visible at the
underhead radius location of the gauge side. The contour plots indicate that with an
increase in the L/V ratio (lateral load), the longitudinal tensile stress increases at the
underhead and gauge corner region. The regions that are in tension can cause a crack to
grow rapidly. The depth at which stress becomes tensile decreases as the L/V ratio
increases. For L/V=0.4, this is of the order of a few mm, as evident from Figure 4.6.
This can result in a transition to Mode I crack growth and the formation of a transverse
defect for rolling contact fatigue initiated cracks that extends to this region and causes
the fatigue crack initiation at the rail underhead radius.
Chapter 4. Underhead radius stresses
95
Figure 4.5 Longitudinal stress distribution with depth on different vertical planes at 16.5, 20, 25 and 35 mm offset from the rail vertical centreline towards the gauge corner at a contact path 30 mm offset and different L/V ratios (0, 0.2, and 0.4)
Figure 4.6 Longitudinal stress (S11) contours on the cross-section of rail with the contact patch offset of 30 mm at midpoint of rail span ( in longitudinal direction at x=2100 mm), during different lateral loads with the L/V ratio of 0, 0.2 and 0.4. The affected region at which the stress becomes tensile shifts up from the rail underhead radius to surface
4.5 Effect of lateral traction direction
As was described in section 4.2, instead of track loading, one rail is modeled to reduce
the computational time and loading cases on high rail are considered only. The direction
of lateral forces will vary between inwards and outwards leading and trailing wheelsets
during curving of conventional three-piece bogie types commonly used in high axle
load railways. In order to understand the effect of the different directions of lateral
traction, loading cases were considered with the lateral traction pointing towards the
gauge side (inward) or towards the field side (outward). To factor these parameters in
terms of the lateral shear traction direction, simulations were conducted for a contact
patch 30 mm offset towards the gauge side from the rail centreline for both inward and
outward shear traction directions with the L/V ratios of 0, 0.2 and 0.4, as shown in
Figure 4.7. The results in Figure 4.7 (a) show high underhead radius stresses even
without inward lateral traction as compared to outward lateral traction. The contact
patch offset and the load cases with an inward / outward force of up to L/V = 0.4 are
just an assumption as a worst condition to predict the resultant stress state at the
underhead locations and this has not been quantified in track measurements. The results
show that by increasing the L/V ratios towards either the field or gauge side of the high
rail, the tensile stress at the underhead radius of the respective side increases. The peak
2
L/V=0 L/V=0.2 L/V=0.4 2
3
1
Chapter 4. Underhead radius stresses
97
tensile stress is about 324 MPa for the inward traction, which is 4 times higher than that
for the outward traction.
Figure 4.7 Mechanical response at the underhead and base fillet with lateral shear tractions (inward and outward); (a) Longitudinal stress variations with different magnitude and direction of shear traction at the underhead offset of measurement point 34 mm along the midpoint (x = 2100 mm) of high rail with contact patch 30 mm offset; (b) Inward lateral deformation (U2) contour of high rail with L/V = 0.4; (c) Outward lateral deformation (U2) contour of high rail with L/V = 0.4
-150
-100
-50
0
50
100
150
200
250
300
350
-0.4 -0.2 0 0.2 0.4
Lon
gitu
dina
l Str
ess
(MPa
)
Towards Gauge Side (Inward) Towards Field Side (Outward)
Underhead Radius Gauge SideUnderhead Radius Field SideBase Fillet Field SideBase Fillet Gauge Side
L/V ratio
Underhead Radius (UHR)
Base fillet
x z
y
High rail (L/V=0.4) High rail (L/V=0.4)
Underhead Radius (UHR)
Base fillet
(a)
(b) (c)
Chapter 4. Underhead radius stresses
98
These stresses are generated by the lateral deformation of the railhead, as shown in
Figures 4.7(b) and (c). It can be seen that the lateral deformation of the railhead for the
inward traction is about 2.5 times higher than that for the outward traction. This is easy
to understand when considering the effect of L and V on the rotation and lateral
deformation of the rail head-on-web. A contact patch offset towards the gauge corner
rotates the railhead counterclockwise in Figures 4.7(b) and (c). The lateral load, L,
causes further rotation and lateral deformation of the rail head. When the lateral load is
towards the gauge corner (inwards), both rotations (due to L and V) are summative and
provide a much larger rotation and lateral deformation than when the lateral load is
outwards. In this case, the rotation due to L partly cancels out the rotation due to V,
producing significantly lower deformation and longitudinal stresses.
4.6 Effect of track foundation stiffness
An important aspect of this study was to model the support characteristics of track
system influenced by operational conditions to investigate the effect of altered stiffness
on the stress state at the rail underhead radius and base fillet. The looseness in sleepers,
deterioration due to ballast pumping and degradation of track bedding would result in
altered stiffness. The altered stiffness may also be due the effect of the looseness in
fasteners and deterioration of rail pads. The model addresses in a very simple way the
effect of altered stiffness. This was achieved by applying varying stiffness values to the
elastic foundation in the lateral and vertical directions of the rail, as described in detail
in chapter 3. This modeling evaluated the stress state at the underhead and base fillet,
when a poor track support condition is presented in a localized area.
The longitudinal tensile stress at the underhead radius and base fillet was investigated
by changing the vertical foundation stiffness underneath the two middle sleepers
adjacent to the contact area. The value of vertical stiffness of all other sleepers was kept
at KV = 1 N/mm3. The value of lateral stiffness was kept at KL = 1740 kN/mm3 for all
the sleepers to achieve a lateral deformation of 1 mm. Figures 4.8 (a) and (b) show the
results of the longitudinal stresses as the foundation stiffness value was altered. The
stress increased at the underhead radius and decreased at the base fillet as the vertical
Chapter 4. Underhead radius stresses
99
Figure 4.8 Longitudinal stress variations at the underhead radius with vertical foundation stiffness at the contact patch 30 mm offset and L/V = 0.4: (a) along longitudinal position of the rail at the underhead offset of measurement point 34 mm; (b) at the underhead offset of measurement point 34 mm and at base fillet at midpoint of rail span (in longitudinal direction x=2100 mm) in between two middle sleepers
-50
0
50
100
150
200
250
300
350
1900 1950 2000 2050 2100 2150 2200 2250 2300
Lon
gitu
dina
l Str
ess
(MPa
)
Longitudinal Position (mm)
0 MN/m35 MN/m123 MN/m
50
100
150
200
250
300
350
0 20 40 60 80 100 120 140
Lon
gitu
dina
l Str
ess (
MPa
)
Vertical Foundation Stiffness (MN/m)
Underhead
Base fillet
(a)
(b)
Chapter 4. Underhead radius stresses
100
foundation stiffness increased from 0 to 123 MN/m, as shown in Figure 4.8 (b). Without
any support to rail at the adjacent sleepers (a vertical foundation stiffness of zero), the
maximum longitudinal tensile stress was found to be 200 MPa at the base fillet and 267
MPa at the underhead, respectively. This condition simulated ballast pumping, which
could remove support from one sleeper, as is observed in the field. Thus, the possibility
of straight breaks at the foot region driven by repeated tensile stresses was enhanced,
becoming much higher. It was noted that beyond the vertical foundation stiffness of 35
MN/m, the longitudinal tensile stress hardly changed at either of the locations. A
comparison of the longitudinal and Von-Mises underhead radius stresses resulting from
variations of support stiffness is also presented in Table 4.1.
Table 4.1 Effect of track support stiffness
Support stiffness
(Vertical)
Longitudinal stress, S11
(MPa)
Von-Mises stress, Mises
(MPa)
Loose support (0 MN/m) 267 283
Average support (35 MN/m) 322 330
Stiff support (123 MN/m)
340 350
4.7 Effect of seasonal temperature variation
The differential temperature due to seasonal effect is another factor which influences the
stress state at the UHR. An approach by the US Department of Transportation [7] for
evaluating the thermal stresses in CWR was adopted here, by taking the service
temperature difference from the neutral temperature. The neutral or stress free
temperature (assuming a temperature condition at which the rail is installed) was taken
as the reference temperature. The two other service temperatures were assumed to
represent the mean of colder months (10 oC), and the mean of the warmer months (42
Chapter 4. Underhead radius stresses
101
Figure 4.9 Longitudinal stress variations at the underhead radius with seasonal temperature at contact patch offset of 30 mm and L/V = 0.4 (a) along longitudinal position of the rail at the underhead offset of measurement point 34 mm; (b) at the underhead offset of measurement point 34 mm at midpoint of rail span (in longitudinal direction at x = 2100 mm) in between two middle sleepers
-500
50100150200250300350400
1900 1950 2000 2050 2100 2150 2200 2250 2300
Lon
gitu
dina
l Str
ess (
MPa
)
Longitudinal Position (mm)
oCoCoC
102642
280290300310320330340350360370
10 15 20 25 30 35 40
Lon
gitu
dina
l Str
ess (
MPa
)
Rail temperature (oC)
Underhead radius gauge side
(a)
(b)
Chapter 4. Underhead radius stresses
102
oC) respectively. The reference temperature was taken as 26 oC and the simulation
temperature was changed for the different service temperatures of the rail. Figures 4.9
(a) and (b) show that the longitudinal tensile stress at the underhead radius increases in
the colder months and decreases in the warmer months with a difference of about 40
MPa. The longitudinal tensile stress went up to 365 MPa at the underhead radius by the
additional tensile stress generated in colder months.
4.8 Fatigue damage
The results of the analyses reveal that the tension spike is strongly dependent on the
contact patch location and the magnitude and direction of the lateral traction, and hence
on the curving and hunting behaviour of the vehicle. Additionally, stiffer tracks
operated in cold weather conditions experience increased underhead radius stresses. The
tension spike at the underhead radius of the rail can cause fatigue damage and cracking.
In addition, the presence of such high tensile stresses can cause a crack to turn
perpendicular to these tensile stresses, resulting in the formation of transverse defects.
In practice, a potential reverse detail fracture could be initiated at sharper curves in the
underhead radius where this localised stress peak has the maximum value. The axle
loads in the Australian heavy haul rail systems are continually rising [89, 139]. This
current analysis, with an axle load of 35 tonnes, shows the propensity for crack
initiation and propagation. With increased axle loads (which sometimes are up to 40
tonnes), these stresses will increase even further and more occurrences of transverse
defect development from rolling contact fatigue damage may be expected.
Vehicle dynamic effects associated with variations in track geometry will also increase
the magnitude of the vertical and lateral loads. For this thesis, the effect of dynamic
loading was assumed to be zero, as the intent was to establish a finite element analysis
approach that reproduced the stress history apparent from the in-situ measured data.
Additional (quasi-static) loads due to curving were included by means of the application
of lateral loads, as described above. Increased train speeds may also produce an increase
in dynamic loads as well as an increase in the effect of lateral forces due to lateral
irregularity on either the gauge or the field side, and subsequently an increase in the L/V
Chapter 4. Underhead radius stresses
103
ratio. This will lead to elevated values of tensile stresses and ultimately an increased
risk of fatigue damage at the underhead radius. In addition, an increase in tensile
bending stresses, together with both residual and thermally induced stresses, could
cause a crack to turn perpendicular to these tensile stresses once they reach a critical
length [3]. With rail uplift ahead and behind the wheel-rail contact, longitudinal tensile
stress develop due to rail vertical bending (reverse bending). A short duration tensile
stress peak occurs due to lateral bending when the instrumented location is directly
beneath the wheel-rail contact.
The underhead region of welded rail joints could be susceptible to such a failure. Welds
have lower material strength, elevated residual stress levels and a complex geometry
leading to stress concentration [42]. The underhead region of the weld collar in
aluminothermic welds often develops fatigue cracking under high axle load conditions
[42]. The fatigue damage will be even more severe when loading is applied on the
gauge side. This is because the tensile stresses increase at the gauge side and underhead
region, as shown in Figure 4.7.
The analysis results provided in this chapter have qualitatively revealed that the
magnitude of the tension spike at the rail underhead radius is strongly dependent on the
contact patch location, the magnitude and direction of the lateral traction, and hence on
the curving and hunting behavior of the vehicle. Rail heads that are heavily worn are
expected to exhibit even higher stress levels, as was identified by Jeong et al in their
examination of reverse detail fractures [6]. The effect of head wear on underhead radius
stresses in relation to the development of transverse defects from rolling contact fatigue
at the gauge corner, is explored in chapter 5.
4.9 Summary
This chapter covered an examination of the stress state at the underhead radius and
gauge corner of the rail, using the commercial finite element package ABAQUS 6.11-2.
The tension spike at the underhead radius was found to be highly dependent on several
factors: the contact patch offset, the ratio of lateral (L) to vertical (V) loads, the
Chapter 4. Underhead radius stresses
104
direction of lateral traction, the vertical foundation stiffness and the seasonal
temperature. It can be concluded from the model results that:
The tension spike at the underhead radius of a rail increases when the contact patch
moves away from the rail centreline and / or the L/V ratio increases. Inward shear
traction was found to be more damaging. The magnitude of longitudinal tensile stress
increases at the underhead radius as the vertical foundation stiffness increases. The
results suggest that the magnitude of longitudinal tensile stress at the underhead radius
increases in the colder months and decreases in the warmer months.
The depth at which stress becomes tensile decreases as the L/V ratio increases. For
L/V=0.4, this is of the order of a few mm. This can result in a transition to Mode I crack
growth and the formation of a transverse defect for both rolling contact fatigue initiated
cracks that extend to this region and the fatigue behavior of the rail underhead, as was
previously examined in connection with the reverse detail fracture in the US.
Further study is also needed to understand the fatigue mechanism and crack initiation at
the underhead location by considering factors such as worn profiles, residual stresses,
and unstable crack growth behaviour at the underhead location, which will be presented
in subsequent chapters.
Chapter 5. Effect of head wear on UHR stresses
105
Chapter 5
Effect of head wear on underhead radius
stresses
5.1 Introduction
In this chapter, the investigation into the effect of head wear on underhead radius
(UHR) stresses is described. The previous chapter qualitatively revealed that the
magnitude of the tension spike at the rail underhead radius is strongly dependent on the
contact patch location, the magnitude and direction of the lateral traction, and hence on
the curving and hunting behavior of the vehicle. The work presented in this chapter
examines the effect of head wear and operating track conditions (the L/V ratio, the
contact patch offset (CPO), foundation stiffness, thermal and residual stress) on
underhead radius stresses.
This study was especially concerned with the qualitative evaluation of the state at the
rail underhead radius for heavily worn rail profiles, which are due to the effect of the
head-on-web bending and the lateral bending of the whole rail profile. It investigated
the situation when the rail head cross-section bends on the rail web. This effect is much
more localized and could be termed local bending stress as a result of vertical and
lateral head-on-web bending. It occurs in the form of a tension spike at the underhead
radius of the rail.
Marich [36-38] has also reported the presence of high tensile longitudinal stresses at the
lower gauge corner of the rail due to localised bending of the head-on-web. The tensile
stress was found to be a function of vertical and lateral load eccentricity and was
Chapter 5. Effect of head wear on UHR stresses
106
influenced by changes in rail head wear (HW). A comparison of different worn profiles
of 60 kg/m and 68 kg/m rail sections under eccentric vertical and lateral loading
revealed that, in terms of allowable rail head wear, the smaller rail section would be
preferable and reported that the bigger rail section was not always better [36]. This did
not take into consideration the increased beam strength with the larger rail, which may
be necessary for the axle loads considered. This work also did not take into
consideration the possible impact of RCF damage; subsequent research by Mutton et al
[1] demonstrated that the approach taken by Marich was therefore non-conservative in
the presence of RCF damage.
Generally, wear results in a reduction of the rail cross-sectional area, and changes the
profile of the rail, which may have an influence on the wheel-rail contact conditions and
consequently impact on the magnitude of bending stresses in the rail. The peak tensile
stress at the underhead radius can be higher in terms of the head wear levels and
potentially produce reverse detail fractures (RDF) (i.e. transverse defects which are
initiated at the lower corner of the gauge face of heavily worn rail) under heavy haul
operations. This type of defect can be observed on heavily worn curved rails on stiff
tracks that are poorly lubricated and subjected to high axle loads, as reported in North
American rail systems [6, 9]. Similar failures have been found in aluminothermic welds
in the Australian heavy haul railway system. Currently, Australian heavy haul rail
systems operate with axle loads of up to 35 tonnes (some recent ones are up to 40
tonnes) [89, 139]. The previous analysis, presented in chapter 4, with an axle load of 35
tonnes considered an unworn rail head. This chapter extends the analysis to consider the
effect of a worn rail profile on longitudinal stresses and their qualitative effect on
fatigue behaviour in the underhead radius of the rail head. The effect of residual stresses
is also incorporated here.
5.2 Finite element analysis model
Several new finite element (FE) models were developed to study the effect of worn
profiles, using the commercial finite elements code ABAQUS (6.11-2). Cases with an
L/V ratio of 0, 0.2 and 0.4 and contact patch offsets 0, 15, 20 and 30 mm were
Chapter 5. Effect of head wear on UHR stresses
107
parametrically studied using six rail profile changes. Six worn rail profiles were adopted
for the case studies, namely: no wear (ideal profile), 5 mm, 15 mm, 20 mm, 22 mm, and
25 mm of head wear (HW), defined as the difference between the highest point on the
new rail and the worn one. The Miniprof unit provides the ability to determine these
values. The rail head profiles used for all head wear (HW) conditions were based on a
typical worn profile selected from recent assessment of rail profiles and rail grinding
activities. These rail profiles assessments were conducted by Welsby et al [140]. The
rail profiles with the different HW used for the current analysis are presented in Figure
5.1. The rail is installed in track at the relevant cant (or inclination) as was shown in
Figure 2.2. The rail inclination is 1:40. Hence the worn profile measurement is not the
vertical dimension at the rail centreline but the vertical dimension with the rail at a 1:40
inclination. However, this does not affect the results or arguments presented in this
thesis as in this work wheel on rail was not modelled, Instead a required Hertzian
contact pressure was simulated on the rail mesh.
15 600M HR - Head wear (HW) of 15 mm for high rail on a 600 m radius curve
Figure 5.1 Worn rail profiles [140]
Ideal Profile
3
4
Chapter 5. Effect of head wear on UHR stresses
108
Figure 5.2 shows a schematic of the modelling approach with worn rail profile. The
detailed positions of the strain gauges for the field measurements by Mutton et al [1] are
also shown in Figure 5.2. The field data used for the model validation was also
discussed in chapter 3. It should be noted that the details on input parameters such as
loading, boundary conditions and support characteristics will remain the same as was
discussed in chapter 3.
Figure 5.2 The model description for the cross-section view of the worn rail (head wear, HW = 25 mm) profile, the strain gauge measurement location and the elastic foundations with variations of contact patch size and shape
Differential temperature due to seasonal effect is another factor which influences the
stress state at the underhead region in continuously welded rail (CWR). The results
presented in chapter 4 revealed that seasonal temperature variation from ambient (26 oC) to cold (10oC) can produce an additional tensile stress of about 40 MPa at the
underhead radius. In order to simulate the stresses resulting from the effect of
Chapter 5. Effect of head wear on UHR stresses
109
temperature variations, the thermal stresses in the longitudinal direction of the rail due
to cold or warm climate were considered. The value of thermal expansion coefficient
was given in Table 3.1.
The input residual stress levels refer to the measurements that were used by
Magiera [136], as discussed in chapter 3. As the rail geometry is quite complex, the
input residual stresses, with six stress components, have to be simplified with uniform
distributions in the proposed rail sections, as shown in Figure 3.14. The input residual
stresses were summarized in Table 3.3 and were applied by defining them as initial
conditions in an input file created in the commercial FE package ABAQUS 6.11-2. For
the first step in the rail FE model calculation, residual stresses became the initial
conditions and no contact loads were applied for this step. It is noted that equilibrium
was reached after some iterations.
5.3 Stress state at the rail underhead radius
The stress state at the underhead radius is influenced by the contact location and the L/V
ratio during the curving and hunting of the vehicle. The work presented in chapter 4
considered in detail the variation of the tension spike and found it to depend on L/V
ratio and contact offset. The tension spike occurs in the rail underhead radius position as
a short-duration spike of about 100 mm in length. The variation in longitudinal stresses
(S11) in this region is worth special consideration as it can play an important role in the
fatigue behaviour of the underhead radius and the propagation of existing gauge corner
cracking. Figure 4.6 shows that an increase in the L/V ratio results in an increase in the
longitudinal tensile stress at the rail underhead radius, and a decrease in the depth at
which the stress becomes tensile. The region with tensile stress shifts up from the rail
underhead radius to the rail surface with an increase in the L/V ratio.
Effect of residual stress 5.3.1
Table 5.1 shows the effect of residual stresses on longitudinal (S11) and Von-Mises
(Mises) underhead radius stresses under different L/V ratios applied at the contact patch
Chapter 5. Effect of head wear on UHR stresses
110
offset of 20 mm for an ideal rail profile. These results reveal that the longitudinal stress
(S11) is reduced by nearly 60 % for L/V = 0, 37 % for L/V = 0.2 and 29 % for L/V =
0.4, if the residual stresses are considered, but the Von-Mises stress (Mises) is increased
at this region with an increase in the residual stress values.
Table 5.1 Effect of stress state at the underhead region due to residual stress
L/V ratio Von-Mises stress, Mises (MPa) S11 (MPa)
with RS without RS with RS without RS
L/V=0 345.2 162.1 38.40 95.5
L/V=0.2 375.4 222.25 121.2 191
L/V=0.4 405.6 282.4 204 286.5
Effect of head wear 5.3.2
Rail head wear results in an increase of the longitudinal stress at the underhead radius.
Figure 5.3 shows the variation of maximum longitudinal stress as the HW changes from
0 mm (no wear) to 25 mm. An increase in the tensile stress is seen from 130 MPa to 350
MPa as the head wear (HW) increases from no wear to a HW of 5 mm. Beyond a HW
of 15 mm, the longitudinal stress changes rapidly from 500 MPa to 940 MPa, and
reaches a maximum value of 940 MPa for a HW of 25 mm. Figure 5.4 shows the
longitudinal stress (S11) distribution in the railhead with different head wear (HW)
cases at an L/V ratio of 0.4 and a contact location of 20 mm offset. In the case of an
ideal profile (Figure 5.4a), compressive stresses are dominant in the rail head and small
values of tensile stresses are seen at the underhead radius with a small area under
tension. With an increase in the head wear, a gradual increase in the tensile stress is
evident with larger tensile values starting at the underhead radius and covering more rail
head area under tension. In addition, the region over which tensile stresses occur can be
seen to have moved closer to the top of the rail.
Chapter 5. Effect of head wear on UHR stresses
111
Figure 5.3 Variation of maximum longitudinal stress (S11) at the underhead region vs rail head wear for an eccentric load located at 20 mm offset from the rail centreline with L/V = 0.4
Figure 5.4 Longitudinal stresses (S11) distribution at the rail head vs the different head wear (HW) profiles for an eccentric load located at 20 mm offset from the rail centreline with L/V = 0.4
0
100
200
300
400
500
600
700
800
900
1000
0 5 10 15 20 25
Lon
gitu
dina
l Str
ess
(MPa
)
Head Wear (mm)
L/V=0.4, Patch offset=20 mm
MPa
Chapter 5. Effect of head wear on UHR stresses
112
The heavily worn profile with HW of 25 mm presents very high longitudinal tensile
stress (S11 > 600 MPa) at the underhead radius and near to the rail head surface, as
shown by the grey region in Figure 5.4(f). The results also indicate that these very high
longitudinal tensile stresses significantly increased up to 946 MPa when the HW = 25
mm, as compared to the other light or moderate head wear cases. The reported 0.2 %
proof stress data for HH (AS 1085.1) rail grade is 780 MPa at standard head position.
The low alloy and some of the HE (Hypereutectoid) rail grades possess the 0.2 % proof
stress ranges to 910 MPa at the standard head position.
Effect of contact patch offset 5.3.3
The position of the contact patch was a key factor under consideration in this study for
two reasons. Firstly, it is known that for curves and a tangent track, the vehicles tend to
exhibit lateral movement or hunting behaviour on the rails. Secondly, the shape and
location of the contact patch changes as the rail head wear profile is influenced by a
reduction in the cross-sectional area of the rail head.
Figure 5.5 Longitudinal stress distribution at the underhead for the different contact locations vs the HW of 0, 22 mm and 25 mm
-100
0
100
200
300
400
500
600
0 5 10 15 20
Lon
gitu
dina
l Str
ess
(MPa
)
Contact patch offset from rail centreline (mm)
HW 0 mm
HW 25 mm
HW 22 mm
Chapter 5. Effect of head wear on UHR stresses
113
Current simulations show that moving the contact location towards the gauge corner
away from the centre of the rail head cross-section causes the tensile stress at the
underhead radius to increase gradually, as shown in Figure.5.5. In order to study the
effect of contact patch location due to the different head wear (HW) levels, a constant
L/V ratio of 0 was taken to minimise the effect of the lateral load (L). In real life the
lateral load will be expected to increase in the presence of vehicle hunting and during
curving (especially combined with traction) and this is investigated in the next section.
Figure 5.5 shows the variation of longitudinal stress as the head wear changes from
HW of 0 mm (no wear) to HW of 25 mm with different contact patch positions. The
results show that, as a result of increasing the contact patch offset, the longitudinal
tensile stress linearly increases for the contact patch offset from 0 to 15 mm. The small
value tensile stress represents a well-maintained rail profile, but the tensile stress
increases significantly after the contact patch offset of 15 mm. The heavily worn profile
results in higher longitudinal tensile stresses at the rail underhead radius when moving
the contact location toward the gauge corner of rail. It can be seen to have peaked at 522
MPa for the case with HW of 25 mm at the contact patch location of 20 mm offset. The
tensile stress was more sensitive to head wear with a 20 mm contact patch offset, which
is representative of poorly maintained rail profiles.
Effect of the L/V ratio 5.3.4
The lateral load (L), which is influenced by vehicle steering or hunting behaviour, was
investigated to evaluate the stress state at the rail underhead radius (UHR). The values
of L were changed to obtain an L/V ratio of 0, 0.2 and 0.4 for this investigation. The
tensile stress generated at the underhead radius was due to the bending of the rail head-
on-web caused by the eccentricity of the vertical and lateral loads from the rail head
centreline. The stress analyses at the underhead radius of the gauge side of the rail
indicated that the tensile stress at the underhead radius position (measurement point
offset of 34 mm) clearly increased with an increase in L/V ratio at the contact patch
offset of 20 mm with respect to different head wear (HW) levels, as shown in Figure
5.6. It was noted that the heavily worn profile showed the worst case, presenting a very
Chapter 5. Effect of head wear on UHR stresses
114
high longitudinal tensile stress of 946 MPa at underhead radius for the L/V ratio of 0.4,
which potentially achieves the yield of the rail material at underhead radius, as shown in
Figure 5.4f. The graph with the no head wear (HW of 0 mm) condition for the tangent
track exhibiting free rolling (L/V = 0) showed a much lower value of tensile stress even
with an increase in L/V ratio up to 0.4, being a 19 % increase in the tensile stress.
However, in the presence of the head wear levels, the tensile stress value became high at
the same location of the underhead radius, almost doubling in value (114 % for HW of
22 mm and 81 % for the HW of 25 mm) in the tensile stress, with an increase in L/V
ratio.
Figure 5.6 Variation of longitudinal stress at the gauge side underhead radius for eccentric loading from rail centreline towards the gauge side for different rail worn profiles versus the lateral traction coefficient
Effect of track foundation stiffness 5.3.5
An investigation was undertaken into vertical foundation stiffness underneath the two
middle sleepers adjacent to the contact position on the longitudinal tensile stress at the
rail underhead radius and base fillet. A low value of vertical foundation stiffness, KV,
presents a poor track support condition. In the previous analysis of the ideal rail profile
(HW of 0 mm), presented in chapter 4, the longitudinal stress at the rail underhead
0100200300400500600700800900
1000
0 0.1 0.2 0.3 0.4
Lon
gitu
dina
l Str
ess
(MPa
)
L/V ratio
HW 25 mm, Patch offset 20 mmHW 25 mm, Patch offset 15 mmHW 22, Patch offset 20 mmHW 0 mm, Patch offset 20 mm
Chapter 5. Effect of head wear on UHR stresses
115
radius increased, but decreased at the base fillet as a result of increasing the vertical
foundation stiffness from 0 to 123 MN/m (see Figure 5.7). Beyond the vertical
foundation stiffness of 35 MN/m, the longitudinal tensile stress became steady for both
locations.
This thesis involved the effect of head wear on longitudinal stresses at the rail
underhead radius and the base fillet under the track support conditions. Figure 5.7
suggests the longitudinal stress increased by 115 MPa and 384 MPa at the base fillet
and underhead radius respectively as a result of increasing the HW from 0 to 22 mm
under average track support stiffness (35 MN/m) conditions, as described in chapters 3
and 4. The increase in the longitudinal tensile stress with HW of 22 mm is more
pronounced at the underhead radius as compared to the base fillet. This is because of
increased local bending stresses at the underhead radius with an increase in the head
wear. This effect has already been discussed in the previous section. The longitudinal
underhead radius stresses resulting from variations of support stiffness for different rail
head wear profiles are also presented in Table 5.2.
Figure 5.7 Longitudinal stress variations at the underhead radius with vertical foundation stiffness at the contact patch 20 mm offset and L/V = 0.4: at the underhead offset of measurement point 34 mm and at the base fillet at the midpoint of the rail span (in longitudinal direction x=2100 mm) in between two middle sleepers
0
100
200
300
400
500
600
700
800
900
0 20 40 60 80 100 120 140
Lon
gitu
dina
l Str
ess (
MPa
)
Vertical Foundation Stiffness (MN/m)
UHR - HW = 22 mm
UHR - HW = 0 mm
Base fillet - HW = 22 mm
Base fillet - HW = 0 mm
Chapter 5. Effect of head wear on UHR stresses
116
Table 5.2 Effect of rail support stiffness on rail longitudinal UHR stresses (S11)
Support stiffness Vertical
Longitudinal stress, S11(MPa)
HW 0 mm HW 22 mm
Loose support
(0 MN/m) 267 599.2
Average support
(35 MN/m) 322 705.8
Stiff support
(123 MN/m)
340 736.6
Evaluation of depth of tensile longitudinal stress 5.3.6
The mechanical response at the underhead radius is significantly influenced when
heavily worn rail profiles are considered. Rail head wear can result in changes to the rail
profile and a reduction in the cross-sectional area of the rail head, both of which can
affect the stress state. The extent to which the loss of a cross-sectional area can be
sustained is related to the direct relationship between the resultant increase in stress
levels and the probability of rail failure. An increase in head wear results in an increase
of the longitudinal stress at the rail underhead radius, as was discussed in section 5.3.2.
As the tension spike occurs in the rail underhead radius position as a short-duration
spike, about 100 mm in length, the variation in longitudinal stresses (S11) in this region
was of special interest as it can play an important role in the fatigue behaviour of the
underhead radius and the propagation of existing gauge corner cracks.
Figure 5.8a shows the longitudinal stress variation with depth for different vertical plane
offsets of 15, 20, and 25 mm from the rail vertical centreline towards the gauge corner,
with an L/V ratio of 0.4 at the contact patch offset 20 mm for HW of 25 mm. By
increasing the vertical plane offset from the vertical centreline of the rail toward the
Chapter 5. Effect of head wear on UHR stresses
117
gauge corner, the longitudinal tensile stress increases, and the depth at which the stress
becomes tensile reduces from 8 mm to 0.7 mm for HW of 25 mm
Figure 5.8 Longitudinal stress distribution with depth on different vertical planes offset from the rail vertical centerline towards the gauge corner, at a contact patch offset of 20 mm L/V=0.4: (a) HW = 25 mm, (b) HW = 15 mm
Plane offset 15 mmPlane offset 20 mmPlane offset 25 mm
(a) HW = 25 mm
(b) HW = 15 mm
Chapter 5. Effect of head wear on UHR stresses
118
For HW of 15 mm (Figure 5.8b), the depth at which stress becomes tensile reduces from
20 mm (plane offset 15 mm) to 0 mm (plane offset 30 mm).
As the contact patch offset can reach a maximum of 25 mm, the longitudinal stress
variation with a depth for different vertical plane offsets of 15, 20, 25 and 30 mm from
the rail vertical centreline towards the gauge corner, with an L/V ratio of 0.4 at the
contact patch offset 25 mm for HW of 20 mm, was examined (Figure 5.9). By
increasing the vertical plane offset from the vertical centreline of the rail toward the
gauge corner, the longitudinal tensile stress increases, but the depth at which the stress
became tensile reduced from 13 mm (plane offset 15 mm) to 0 mm (plane offset 30
mm) for HW of 20 mm as shown in Figure 5.9.
Figure 5.9 Longitudinal stress distribution with depth on different vertical planes at 15, 20, 25 and 30 mm offset from the rail vertical centerline towards the gauge corner at a contact patch offset of 25 mm L/V=0.4 HW 20mm
A comparison of the stress variation with realistic head wear is illustrated by contour
plots (Figure 5.10). The line shows the tension / compression boundary and a
subsequent upward shift in this boundary as a function of increasing head wear. Figure
5.10 also shows the longitudinal stress distribution as a function of depth on a vertical
plane at a 25 mm offset from the rail vertical centreline towards the gauge corner for
contact patch offsets of 20 mm with L/V 0.4 with different head wear (HW). The high
head wear (HW) results in an increase of the longitudinal stress at the underhead radius
(UHR) and the region of tensile stresses moves closer to the top of the rail. These results
illustrate the severity of longitudinal tensile stresses at the underhead radius (UHR)
region with increasing head wear (HW) conditions.
Figure 5.10 Longitudinal stress distribution (S11) with depth on a vertical plane at a 25 mm offset from the rail vertical centerline towards the gauge corner for contact patch offsets of 20 mm with L/V 0.4
Chapter 5. Effect of head wear on UHR stresses
120
Figures 1.2, and 1.3 [2] show the development of a crack at a gauge corner in a worn
rail, its turning down in the rail head and the final fracture as a transverse defect. Such
high tensile bending stress could help progress transverse defects (TDs) from the RCF,
which generally start to turn down at about 5 mm below the rail head surface, as
examined by Mutton et al [1, 2]. Another example was given by Mutton et al [2],
indicating that the RCF cracks at the gauge corner turn down in low alloy heat treated
rail grades as well.
The crack is expected to turn down and grow rapidly when it enters the region of tensile
bending stresses. Its growth will depend on the magnitude of the tensile stress as
compared to the contact stress field, material properties such as anisotropy etc. In the
case of an HW of 25 mm, the zone of tensile bending stress was very close to the
contact surface (Figure 5.10), which could cause a rolling contact fatigue crack to turn
perpendicular to the running surface, resulting in the formation of transverse defects due
to Mode I, as has been observed in practice [1, 2]. The presence of a high underhead
radius stress peak in a heavily worn rail head is evident, as is shown in Figure 5.10. In
this case, the Mode I transition of a pre-existing RCF crack to form transverse defects
has been caused by local bending rather than global bending.
5.4 Summary
This chapter has described the stress state at the rail underhead radius, influenced by the
head wear (HW) levels in conjunction with the wheel-rail contact conditions and / or
L/V ratio. The rail underhead radius in a heavily worn rail presents the most critical case
with very high tensile stress close to the contact surface. The wheel-rail contact location
is another critical factor that influences the stress state at the rail underhead radius,
particularly when moving laterally towards the gauge side. Lateral loading generated by
curving and hunting operations can also be highly influential on the stress state at the
rail underhead radius. Stiff tracks are more prone to high underhead radius stresses.
The depth at which tensile longitudinal stresses occurred was evaluated and it was also
demonstrated that residual and CWR stresses interact with these high local bending
Chapter 5. Effect of head wear on UHR stresses
121
stresses, and can potentially initiate a fatigue crack or cause existing rolling contact
fatigue cracks to turn downwards and form transverse defects. In addition, the results
presented in this chapter have helped to understand the mechanism of rail damage
qualitatively.
This page is left intentionally blank
Chapter 6. Fatigue damage prediction
123
Chapter 6
Fatigue damage prediction
6.1 Introduction
This chapter examines the possibility of fatigue damage initiation at the rail underhead
radius (UHR) due to the occurrence of a short duration tensile stress peak when a wheel
directly passes over this region. Heat-treated low alloy, euctectoid and hypereutectoid
rail grades operated under heavy haul conditions are considered. It was demonstrated in
chapters 4 and 5 that tensile stress generated at the underhead radius is mainly due to the
localised bending of the rail head-on-web induced by the complex wheel-rail contact
conditions that are associated with a combination of lateral offsets in the contact
positions and lateral (transverse) forces. In the presence of heavily worn rail, the
magnitude of these stresses could be even higher. The main focus of this chapter has
been an evaluation of the potential risk of fatigue damage at the underhead radius
(UHR) resulting from a combination of wheel-rail contact conditions and rail head wear
(HW) states.
Mechanical responses at the underhead radius have been explored using the finite
element method (FEM). The potential fatigue mechanism and crack initiation at the
underhead radius are described in this chapter as damage parameters. The Dang Van
(DV) criterion, implemented as a customised computer programme, was used to identify
any potential fatigue damage at a specified location. The Palmgren - Miner law was
used to quantify damage accumulation and potential cycles to failure. Fatigue behaviour
was compared for the different high strength rail steel grades used for heavy haul
operations in order to predict rail wear limits.
Chapter 6. Fatigue damage prediction
124
Rail wear can result in changes to the rail profile and a reduction in the cross-sectional
area of the railhead, both of which can affect the stress state under in-service loading
conditions. The extent to which a loss in the cross-sectional area can be sustained is
related to the direct relationship between the resultant increased stress levels and the
probability of rail failure. Wear itself may therefore be a direct threat to rail integrity,
but excessive wear combined with the presence of a defect that increases the risk of
fatigue cracking may pose an even greater risk [48]. Wear limits for rails can therefore
be determined by examining the relationship between rail wear and stress levels
throughout the rail section, and the probability of fatigue damage either in defect-free
rail, or in the presence of known defect types. Defect types that may be considered
include rolling contact fatigue (RCF) and reverse detail fractures (RDF). Additional
defect types may be associated with rail welds such as those produced using flash butt
or aluminothermic welding procedures.
Marich [36] reported that rail wear under high axle load conditions can generally be
observed on the gauge face of the high rail of curves with radii less than 500-800 m due
to the increased lateral loads and the occurrence of wheel flanging. Loss of material at
the running surface of the rail also occurs due to a combination of wear resulting from
normal wheel/rail interaction, and rail maintenance activities such as rail grinding. The
influence of increasing rail head wear levels on rail stress levels can be examined
numerically using finite element analysis. Marich [38] also measured rail stress levels
under both laboratory and field conditions to define acceptable rail wear limits for high
axle load conditions, based on the fatigue behaviour of the rail material. It was found
that rail wear limits could be considerably greater than what was currently accepted in
practice.
In a rail section, there are two deformation zones, which can be simply classified in
terms of the mechanical responses caused by the wheel-rail contact conditions. The first
zone, which is close to the contact surface, is the region of plastic deformation due to
repeated rolling/sliding contacts between the wheels and the rail. As a consequence, if
the contact loads are high, crack initiation and propagation in the contact region leading
to rail defects, such as rolling contact fatigue and squats, could be very significant. This
Chapter 6. Fatigue damage prediction
125
behaviour is of great significance to the rail industry due to the increased risk of rail
failure, and has therefore been the subject of much research [48, 82-85]. The hypothesis
for these studies is based on the theory of plasticity with a suitable fatigue criterion for
the rail life prediction. The second zone is the underhead region, which is away from the
contact surface but is influenced by the same wheel-rail contact conditions. To predict
the initiation of a crack in this region, the Dang Van fatigue criterion can be used, as it
has previously been used in many industrial applications, particularly in the automotive
industry [28, 52]. The fatigue criterion is based on a multi-scale approach and on a
shakedown limit hypothesis. As the wheel-rail contact loads cause a multi-axial state of
stress with out-of-phase stress components and varying principal stress directions, such
a fatigue criterion is suitable and is thus used to predict fatigue limits and crack
initiation in a rail for this thesis [51]. Figure 6.1 shows the results for in-depth analysis
of the stress components.
Figure 6.1: In-depth stress analysis at underhead offset of 25 mm for head wear of 22 mm, contact patch offset = 15mm, L/V=0.4
-100
0
100
200
300
400
500
600
1900 1950 2000 2050 2100 2150 2200 2250 2300
Valu
e of
Sre
ss (M
Pa)
Distance x-position (mm)
sxx
Txz
Syy
Szz
Txy
Tzy
Smises
25mm
Chapter 6. Fatigue damage prediction
126
In sharper curves, as the curving forces increase, the deformation of the rail material
potentially results in a flow lip at the lower gauge corner. With the continued passage of
rail traffic, a fatigue defect may be initiated and propagated into the rail head [9]. The
same behaviour is considered significant in reverse detail fractures (RDF), which are
found to occur in North American rail systems [6-9] due to poorly lubricated, heavily
worn rails on stiff tracks subjected to high axle loads. The potential for reverse detail
fractures is higher in sharper curves, due to the increased probability of the localised
stress concentrations associated with heavy gauge face wear and plastic flow. Reverse
detail fracture is also influenced by longitudinal bending stresses at the underhead
radius, as reported by Jeong et al [6]. Failures due to fatigue cracking at the underhead
radius of aluminothermic welds under heavy axle load conditions were examined by
Salehi et al [42]. The welds exhibited an increased sensitivity to this behaviour due to a
combination of elevated residual stress levels, a more complex local geometry leading
to increased stress concentration, and differences in material characteristics compared to
the parent rail.
Some of the above aspects were examined in chapters 4 and 5. The current chapter
examines the probability of fatigue damage in the rail underhead radius, under the
influence of localised tensile stresses and in the absence of pre-existing defects, such as
transverse defects and reverse detail fractures. These aspects were examined in the
context of heavy haul (high axle load) conditions, based on the behaviour of 68 kg/m
rail. The influence of rail material grade was considered, based on the characteristics of
a number of heat-treated low alloy and hypereutectoid grades commonly used under
these conditions [2, 13, 130]. A high cycle fatigue analysis based on Dang Van criteria
[51] was used to examine the fatigue behaviour for a range of wheel-rail contact
conditions and varying amounts of rail wear. In addition, crack propagation was
simulated using the Pamgren - Miner damage accumulation law [62, 103], which can be
used to determine the extent of material damage at the underhead region and predict the
number of cycles to failure at a specified reference point.
Chapter 6. Fatigue damage prediction
127
6.2 Model development
The finite element (FE) models developed in chapters 3 and 5 to calculate a stress state
were extended to evaluate fatigue damage around the reference point set at the
underhead radius. The Dang Van criterion was programmed into FORTRAN-code and
implemented into ABAQUS through UVAR-M (User output variable), to identify any
potential fatigue damage. A wheel load was applied on the rail, assuming a fully
slipping Hertzian contact pressure, at the midpoint of the rail span (in longitudinal
direction at x = 2100 mm) in between two middle sleepers. Different contact patch
offsets (0, 15 20 and 30 mm) were considered. A contact patch offset of 20 mm is
shown in Figure 6.2. It should also be noted that the details on input parameters such as
loading, boundary conditions and support characteristics will remain the same as were
presented in chapter 3.
In real life, the whole stress cycle at a point is obtained as the contact loading moves
from one end of the rail span to the other, as shown in Figure 6.2a (not to scale), with
initially a low stress which becomes maximum when the wheel contact patch is on top
of the reference point under consideration and returning to low once the wheel has
passed the reference point. As the rail geometry being consistent for the entire rail span,
this stress cycle can be simulated by keeping the contact load fixed on one point but
moving the point under consideration from one end to the other.
A separate study was conducted for fatigue analysis at the rail underhead radius
throughout the span (S = 4200 mm, Figure 6.2b, not to scale), to check the value of
maximum fatigue damage under given extreme loading conditions. The maximum
fatigue damage was found to occur at the underhead radius, when the wheel load on the
rail surface was close to the position of interest. This approach was found to be
sufficiently good to perform the study. Therefore, fatigue damage analysis was
performed for a very fine mesh of 30 mm width of rail length in a longitudinal direction
at the underhead radius, with respect to the contact patch of 20 mm width (semi major
axis a = 10 mm) along the longitudinal direction of the rail, to accurately capture the
value of the maximum fatigue damage and reduce the computational time, as shown in
Chapter 6. Fatigue damage prediction
128
Figure 6.2c. Coarse mesh was used for the rest of the rail span (S = 4200 mm). In the
parametric studies, rail profiles representing no wear (ideal profile), 5 mm, 10 mm, 15
mm, 20 mm and 22 mm of head wear (HW) were used. The finite element model
construction, the required parameters and the modelling validation have already been
discussed in detail in chapters 3 and 5.
Figure 6.2 Representation of model setup for DV fatigue damage analysis
(a) (b)
(c)
Chapter 6. Fatigue damage prediction
129
Dang Van fatigue criterion 6.2.1
The Dang Van (DV) criterion is a shear stress based criterion that is applicable for stress
levels below the elastic shakedown limit of the material. If the following inequality is
satisfied on a shear plane passing through each material point at least once in the whole
stress cycle, damage occurs. This inequality is expressed as:
( ) ( ) (6.1)
Ļa(t) is the time dependent value of shear stress on the specified shear plane at the
specified material point and is defined as the difference between the instantaneous and
mean shear stress of the loading cycle; Ļh is the time dependent hydrostatic stress at the
material point. The constants (aDV and bDV) are the functions of the material fatigue
limits [28]. The fatigue limits f-1 and t-1 can be obtained from the classical experimental
bending and twisting tests respectively, and are stated in Table 6.1 based on estimated
fatigue properties. The calculations are shown in the next section. The constants are
calculated as:
( )
(6.2)
The value of the inequalityās left side represents a numerical index for fatigue damage,
taken as the DV (Dang Van) damage parameter applied to the dense mesh area at the
underhead radius (see Figure 6.2c) for subsequent investigations.
Table 6.1 Material properties and constants of rail
Implementation of the critical plane approach 6.2.2
The plane on which the above-mentioned inequality is satisfied is called the critical
plane, as shown in Figure 6.3. However, the critical plane was not obvious at the
beginning of the analysis so the inequality needed to be assessed in all shear planes
passing through each material point being investigated for potential fatigue crack
initiation. The critical plane and DV damage parameters were calculated using the
corresponding algorithms and equations implemented into customised FORTRAN-
code, linked to the FE package ABAQUS, UVAR-M (user output variables) as given in
Appendix C. The Dang Van criterion was programmed based on equations presented by
Ekberg et al [62]. The validity of the criterion under various stress cycles and the mid
value of shear stress have been described elsewhere in the literature [51, 103, 141]. The
implementation in ABAQUS subroutine āUVAR-M was further checked by modeling a
shaft made of ductile material. In torsional loading, it will fail along the plane of
maximum shear, that is, a plane perpendicular to the shaft axis. The list of defined
UVAR-Mās for the output variables is given in Table 6.2. A description of UVARs 2 to
5 on a critical plane is given in Figure 6.3.
Table 6.2 Definition of UVARs for the output variables
UVARs Meaning
UVAR1 The inequalityās (Equation 6.1) left side, EQ, (MPa) UVAR2 Angle between the x axis and the external normal, n, of the critical plane
UVAR3 Angle Ī² between the y axis and the external normal, n, of the critical plane
UVAR4 Angle between the y axis and the external normal, n, of the critical plane
UVAR5 The shear stress at the critical plane, (MPa)
UVAR6 The shear stress amplitude, a(t), (MPa)
UVAR7 The damage parameter, Di, at the ith number of the equivalent stress cycles
UVAR8 The predicted number of cycles to failure, Nf
Chapter 6. Fatigue damage prediction
131
Figure 6.3 Description of UVARs 2 to 5 on a critical plane
Estimated fatigue properties of different rail grades 6.2.3
For high axle load rail operations, high strength, heat-treated rail grades are commonly
used, as these provide increased resistance to damage in the form of wear and rolling
contact fatigue [2]. During manufacture, heat treatment is carried out using either in-line
or off-line processes to increase hardness levels throughout the rail head. In addition,
steel compositions may vary from eutectoid carbon-manganese grades to low alloy
grades and more recently hypereutectoid grades with carbon contents up to 1% [130].
Hardness levels throughout the head of these rail types decrease with increasing depth
below the outer surface. The effective depth of hardening is typically greater with in-
line processes, while hardness levels are generally higher in the low alloy and
hypereutectoid grades. For this thesis, rails representing the three categories described
above were used. The variation in hardness along a traverse from the gauge corner to
the intersection of the fishing surfaces in these grades has been examined previously [2,
13]. The material grades and typical compositions are summarised in Table 6.3.
O
Chapter 6. Fatigue damage prediction
132
Table 6.3 Chemical composition [43, 130]
Rail Type C %
Si %
Mn %
Cr %
P %
S %
HH 0.65 0.58 1.25 0.2 0.025 0.025
LAHT 0.72 0.13 0.65 0.6 0.025 0.025
HE 1.0 0.55 0.7 0.3 0.025 0.025
The material fatigue limit, as defined by the stress at which failure occurs after a
specified number of loading cycles, is a key factor when describing rail life. The fatigue
limit under rotating bending, f-1, is proportional to the ultimate tensile stress (Y.) of the
rail, with a proportionality ratio in the range 0.25-0.6 [142]. For the present study, a
ratio of 0.38 at 106 cycles was used for the underhead region, while the twisting fatigue
limit, t-1, was set to 58% of the f-1 [142]. Generally a ratio of 0.5 is considered suitable
for steel. However there should be a reduction for surface roughness, which would give
a ratio lower than 0.38. Since the gradients are high, this value could probably be
increased. Three fatigue limits with different rail grades are provided to observe the
sensitivity of these limits in detail. The mechanical properties of relevance are for the
underhead region, and hence may differ from those measured at other locations in
accordance with the relevant specifications. The value of the ultimate tensile stress in
this region can be estimated from the hardness according to reference [13].
Experimental results for hardness distribution 6.2.4
Figure 6.4 shows the hardness testing apparatus used. Hardness measurements were
carried out in the underhead radius region of the samples (specimen in Figure 6.5)
prepared from the above mentioned rail grades, and these results are plotted in Figure
6.6 and were used to estimate the fatigue properties summarised in Table 6.4.
Chapter 6. Fatigue damage prediction
133
Figure 6.4 INDENTEC-Hardness measurement apparatus (Courtesy of Swinburne University of Technology)
Figure 6.5 Rail sample specimen (section and the shape) for hardness testing
Chapter 6. Fatigue damage prediction
134
Figure 6.6 Hardness distribution along transverse plane (underhead radius measurement point offset 20 mm) of high strength rail steels
Table 6.4 The estimated fatigue limits at the UHR for different high strength rail grades
Rail grades Hardness (HV)
Y (MPa)
f-1 (MPa)
t-1
(MPa)
HH 270 615 353 205
LAHT3 320 753 424 246
HE3 370 895 480 278
Damage Accumulation 6.2.5
The DV criterion employed in the model was used to identify the damage accumulation
that occured during a specific time portion of the analysis when the inequality (6.1) at
each material point was fulfilled. To quantify the damage, the Palmgren-Miner linear
damage accumulation rule [103, 143] in conjunction with the Wƶhler curve [62, 143]
250
270
290
310
330
350
370
390
410
430
450
0 2 4 6 8 10 12 14 16 18 20
Har
dnes
s (H
V30)
Distance from centre of railhead cross-section (mm)
HE3 (HE-X)LAHT3HH (AS1085.1)
Chapter 6. Fatigue damage prediction
135
was also used. The Wƶhler curve defines a linear relationship between fatigue stress and
number of cycles to failure in which the fatigue threshold stress is in a range of t-1
corresponding to failure at the t-1 for 106 stress cycles (the highest cycle) and at the DV
magnitude of the fracture stress for 101 stress cycles (the lowest cycle), as shown in
Figure 6.7. For this thesis, the stress cycle for each case was evaluated to ensure that
there was no plasticity, a basic condition of the DV criterion. It should also be noted
that, in the current thesis, the Wohler curve, Figure 6.7, is used only to demonstrate the
methodology. The loading in this case is non-proportional, for practical implementation,
it would be desirable to use material properties correlated with extensive experimental
and site observations that represent the in-service loading conditions.
Figure 6.7 The fatigue threshold stress is in a range of t-1 corresponding to failure at the t-1 for 106 stress cycles (the highest cycle) and at the DV magnitude of the fracture stress for 101 stress cycles (the lowest cycle)
In terms of the Palmgren - Miner linear damage accumulation rule, the damage
degradation can be determined through the following expression:
Where, Di: is the damage corresponding to the ith equivalent stress cycles,
i: is the number of the shear stress cycles
ĻEQi: is the ith equivalent stress calculated by Equation 6.1
In Equation 6.3, at the low stress cycles of 101, the damage parameter, D, is equal to 0.1
when the reaches the bending fatigue limit, f-1. At the high stress cycles of 106, the
damage parameter, D, is equal to 10-6 when the ĻEQ reaches the twisting fatigue limit, t-1.
The damage degradation at each material point via the ith stress cycles to failure, Nf, can
be defined as:
Di = 1/Nf. (6.4)
6.3 Residual stresses
The overall stress state in rails under service loading conditions can also be influenced
by the residual stresses. Therefore, residual stresses in the rail are an important factor
contributing to the overall stress state, and hence some rail specifications include upper
limits for residual stress levels [81, 127]. The distribution of residual stresses in as-
manufactured rails is influenced by straightening procedures such as roller-straightening
[144]. The measurement of residual stresses in rails is usually undertaken using a
destructive method involving strain gauges [26]. This method does not provide
information on the distribution of residual stress at depths below the outer surface,
which can be determined using neutron diffraction [145].
In previous studies of RCF crack initiation and propagation, there was rare
consideration given to residual stress scatter in the subsurface for most models used [27,
120, 146]. Since fatigue life is dependent on the weakest point of a material, the local
extreme is most relevant for the fatigue life prediction. For the current study, the
residual stress distribution was based on the measurements conducted by Magiera [147].
As the rail geometry is quite complex, the residual stresses were considered in terms of
the six stress components for each of the regions shown in Figure 3.14. The input
residual stresses are summarized in Table 3.3.
Chapter 6. Fatigue damage prediction
137
6.4 Plasticity of the underhead radius
The stress state in the underhead region needs to satisfy the basic conditions of the DV
criterion, which states that stress during the loading cycle remains elastic. The
maximum value of Von-Mises stress, Mises max, at the underhead region was calculated
for L/V ratio of 0, 0.2 and 0.4, contact patch offset (CPO) values of 15 and 20 mm and
different rail worn profiles with a maximum of 22 mm of head wear.
Figure 6.8 Von Mises stress distribution at the rail underhead radius vs different head wear profiles: for eccentric loads located at different offset from the rail centerline with L/V = 0, 0.2, 0.4
The yield region is defined by Mises max > Y. The yield stress ( Y) values of three high
strength rail grades were used in this analysis. The yield stress ( Y) value used were for
a plain C-Mn Head Hardened grade, HH (AS 1085.1) - Y = 615 MPa [80], Low Alloy
Heat Treated grade, LAHT3 - Y = 753 MPa [127] and Hypereutectoid heat treated
grade HE - Y = 895 MPa [81] at the underhead radius. These values were estimated
0100200300400500600700800900
1000110012001300
0 2 4 6 8 10 12 14 16 18 20 22
Von
Miss
es S
tres
s -M
ises (
MPa
)
Head Wear (mm)
L/V=0.2, Patch offset=20 mmL/V=0.4, patch offset=15 mmL/V=0, Patch offset=15 mm
HE3- ĻY
LAHT3- ĻY
HH (AS1085.1)- ĻY
Yield Region
Increasing the yield strength
Chapter 6. Fatigue damage prediction
138
based on hardness testing conducted at the underhead radius, as reported in Table 6.4.
The yield stress values at the underhead region for the HH, LAHT and HE3 rail grade
are also illustrated in Figure 6.8 by three dashed lines, which are used to examine the
yield criteria for different materials. The effect of increasing the yield strength of the
rail material for the same range of loading conditions is also shown in Figure 6.8.
Increasing the yield strength enables the limit to be increased for high strength rail
material grades.
The results (Figure 6.8) indicated that, for a heavily worn rail profile, the Von-Mises
stress at the underhead region increased. Under the severe loading case (CPO = 20 mm,
L/V = 0.2), the Von-Mises stress remained below the yield stress of the HE3 and
LAHT3 rail grades up to the maximum head wear considered. For the HH rail grade, the
Von-Mises stress exceeded the yield stress at a head wear value of 18 mm, for an L/V
ratio of 0.2 applied and CPO of 20 mm from the rail centerline; reducing the CPO to 15
mm at an L/V ratio of 0.4 increased this limit to 22 mm. For HH rail, above these head
wear limits implies (global) plastic flow and thus the DV criterion is not applicable.
6.5 Dang Van (DV) damage parameter
The model uses a wheel load of 172 kN, which is equivalent to an axle load of 35
tonnes. The L/V ratios of 0.2 and 0.4 represent loading conditions in curves of
decreasing radii. For tangent track, an L/V ratio of 0 is used, although this may increase
slightly in the presence of vehicle hunting. The curved track loading cases also include a
contact patch offset of 15 mm from the centre of the running surface for a well
maintained rail profile, and 20 mm for poorly maintained rail profiles, as was reported
by Marich [36].
Fatigue failure is predicted to occur at the rail underhead region when the DV damage
parameter exceeds the torsional fatigue limit t-1. Figure 6.9 shows the extent of the
region at the underhead radius, which satisfies this criterion under the loading cases
examined for the HH rail grade. For a contact patch offset of 15mm and an L/V ratio of
0 (Figure 6.9(a)), fatigue damage is first predicted to occur at a head wear of 20 mm; the
Chapter 6. Fatigue damage prediction
139
region affected increases in size for a head wear of 22mm. For a tight radius curve,
represented by a contact patch offset of 20mm and L/V ratio of 0.2, fatigue damage is
first predicted at 14 mm head wear. The critical plane on which the fatigue damage
occurs at point A in Figure 6.9 (b) is at 108O in the x-axis, at 144O in the y-axis and at
60O in the z-axis, respectively. The angles of the critical plane are defined in Table 6.2
and its directions shown in Figure 6. 2.
(a) Contact patch offset 15 mm with L/V=0
(b) Contact patch offset 20 mm with L/V=0.2
Figure 6.9 Variation of fatigue damage with rail head wear at the underhead radius for an eccentric load located at (a) 15 mm offset from the rail centerline with L/V=0; (a) 20 mm offset from the rail centerline with L/V=0.2, for a plain C-Mn Head Hardened (HH) grade [80], with t-1 = 205 MPa
HW = 22 mm HW = 20mm HW = 14 mm
HW = 18 mm HW = 14 mm
HW = 10 mm
Point A
UVARM1 (MPa) - DV Damage Parameter
Chapter 6. Fatigue damage prediction
140
In practice, the head wear (HW) limit of rail has been stated by Duvel et al [13] to be in
the range of 20 to 15 mm at the curves of decreasing radii. The current head wear limit
is set to the 22 mm for tangent track and 15 mm for tight rail radius curves reported for
heavy haul railways in Australia [33, 148]. Typically, HW = 20 mm is allowed for
Deutche Bahn (DB), as reported by Zerbst et al [124]. Table 6.5 shows that the
predicted head wear limits are in line with practice under the proposed extreme loading
cases, but the predicted value may be conservative. However, by considering the
loading conditions with the lateral shear traction in the outward direction, Marich [38]
found that the acceptable wear limit would be 27 mm for 600-800 m radius curves at an
L/V ratio of 0.3 for rail material with a fatigue strength of 240 MPa. The HW drops to
20 mm if an additional tensile stress of 80 MPa due to variations in rail temperature
below the stress-free or neutral temperature were considered. Both of these values are
greater than what is currently accepted within the rail industry.
Table 6.5 Approximate comparison of predicted head wear limits for head hardened rail based on fatigue limits with extreme loading cases for different track conditions compared to current head wear limits [124, 148]
Rail track curvature
Rail profile maintenance
Assumed L/V ratio
Patch offset (mm)
Predicted HW limit (mm)
Current HW limits (mm)
Tangent rail track
well 0 15 20 22
poor 0 20 15
Tight radius curve
well 0.4 15 14 15
poor 0.2 20 13
Figure 6.10 shows the variation in Dang Van damage parameter with an increase in
head wear for a range of loading conditions and the three rail grades considered. The
fatigue prone region for the three rail grades is represented as horizontal lines for t-1 of
HH, LAHT3 and HE respectively, as shown in Figure 6.10. For the HH rail grade,
Chapter 6. Fatigue damage prediction
141
initiation of fatigue damage is predicted at approximately 14 mm of head wear at an
L/V of 0.2 and contact patch offset (CPO) of 20 mm, or an L/V ratio of 0.4 and 15 mm
of contact patch offset. The corresponding limit for an L/V ratio of 0 and contact patch
offset of 15mm is approximately 20 mm. However, with an increased fatigue strength
there is also an increased sensitivity to surface roughness, Figure 6.10, shows the
estimated fatigue limits for three different rail grades to see the sensitivity of these
grades against fatigue crack initiation under given loading conditions. Increasing the
fatigue strength enables the head wear (HW) limit to be increased for high strength rail
material grades.
Figure 6.10 Variation of DV damage parameter versus rail head wear HW at the underhead radius for an eccentric load located at different offset from the rail centreline with L/V=0, 0.2, 0.4, for three rail grades: a plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat Treated (LAHT) grade [127], and Hypereutectoid (HE) Heat Treated grade [81]
Wohler curve used in the current analysisHW 22 mm, L/V=0.2, Patch offset=20 mmHW 22 mm, L/V=0.4, Patch offset=15 mmHW 22 mm, L/V=0, Patch offset=15 mmHW 14 mm, L/V=0.2, Patch offset=20 mmHW 14 mm, L/V=0.4, Patch offset=15 mm
Fatigue Region
Chapter 6. Fatigue damage prediction
145
corner on heavily worn rail. Excessive wear combined with a pre-existing transverse
defect may pose a direct threat to rail integrity. The treatment of residual stress aspects
needs to consider the fact that they may differ from those used for the previous
measurements and may not be representative of current rail manufacturing procedures.
Although the analytical results indicate under what combination of rail grade, rail head
wear and loading conditions fatigue damage is predicted to initiate, in practice fatigue
failure is generally associated with the presence of pre-existing defects or a stress
concentrator in the form of a sharp radius which may include flow lips (as was evident
in the reverse detail fracture mode) or rolling contact fatigue cracking. Material defects
are not included but different rail material grades are studied. In the case of defects, this
provided an overview assessment and is deemed sufficient.
6.7 Summary
The fatigue damage prediction analysis for the rail underhead radius was undertaken to
investigate fatigue damage behaviour as a function of the grade of rail material and
increasing head wear under heavy haul operations. Fatigue damage was predicted using
the Dang Van criterion. The analyses revealed that under the more severe loading
conditions, fatigue damage can be predicted to develop at the rail underhead region with
increasing amounts of head wear for the HH rail grade. For rail grades with high fatigue
limits, fatigue damage rarely occurred at the underhead region under the range of
loading conditions considered with an increase in head wear. Lateral offsets in the
wheel contact patch relative to the rail centreline, in combination with increased lateral
forces associated with vehicle curving, increase the probability of fatigue damage
initiation at the underhead radius. This is due to an increase in the tensile stresses at the
underhead region, which is induced by the local bending behaviour on the rail web. In
the presence of heavily worn rail profiles, the fatigue life of rail in sharp curves is less
than in tangent track. In general, the more wear that occurs on the surface of rail, the
more fatigue damage can be predicted to initiate at the rail underhead radius.
Chapter 6. Fatigue damage prediction
146
The approach presented in this chapter, if extended to include the residual stress aspects
outlined above and in the presence of defects, provides a basis for an assessment of
allowable rail wear limits as a function of loading conditions and the rail material grade.
Further work is required to examine these aspects of rail fatigue damage behaviour.
Chapter 7. Rapid fracture modelling using X-FEM
147
Chapter 7
Rapid fracture modelling using X-FEM
7.1 Introduction
The results presented in chapters 4 and 5 have shown that the occurrence of tension
spikes as a result of lateral head bending and localised vertical and lateral bending of the
head-on-web are significant to the understanding of the unstable propagation of pre-
existing RCF cracks. This chapter investigates the growth behaviour of long RCF
cracks, in particular, the tendency of rail to break due to rapid fracture, under the high
axle load conditions typical of those that exist in Australian heavy haul operations. This
work is also very interesting from the perspective of focusing on the impact of tension
spikes on the unstable growth behaviour of gauge corner cracking (GCC). This
behaviour is exacerbated with increasing rail head wear (HW), and the propensity of
rapid fracture associated with this behaviour correlates with the extent of rail head wear.
An investigation of fatigue crack initiation in the rail head and the fatigue crack
propagation in the web of the weld zone of a rail was conducted by Josefson et al [149],
using multi-axial fatigue analyses incorporating Dang Van theory. The effect of global
rail bending stresses was incorporated in this model by Ringberg et al [150]. A fatigue
crack growth model developed by Skyttebol et al [151] was used to examine the crack
growth behaviour of a welded rail with respect to the influence of residual stresses.
These investigations focused on the crack growth in phases 2 and 3 of āWhole Life Rail
Modelā. It was observed that a weld is more sensitive, or less fatigue resistant, to
surface cracks than to embedded cracks. As expected, the reason found was that stress
ranges in the location of surface cracks were larger than for embedded cracks. Nielsen
Chapter 7. Rapid fracture modelling using X-FEM
148
et al [152] also found the same results in their work and suggested that surface defects
should be avoided. The numerical investigations demonstrated that crack growth rate is
influenced by impact loads and by seasonally-dependent thermal stresses. It was found
that these factors also have a great effect on the fatigue life of a rail.
The risk of rail breaks was investigated by Sandstrƶm et al [111] from a mechanical and
statistical point of view. In particular, the influence of impact loads from flatted wheels
was considered. Wheel flats were not considered to contribute to an increase in rail
bending even though they increased the contact forces. Hence, the influence on fatigue
crack growth was minor. However, bending stress was a major concern regarding any
final fracture. Kapoor et al [18, 23] and Dutton et al [21] developed āWhole Life Rail
Modelā (WLRM), during their investigation of the Hatfield accident, as was described
in detail in the introductory chapters. A phase 3 crack growth dominated by bending
stresses has a significantly higher crack growth rate. A rapid fracture resulting from
transverse defects (TDs) such as shown in Figure 1.3, may occur if a RCF crack turns
down, additionally driven by the tensile bending stress.
This chapter examines rapid fracture from a large crack at the gauge corner of rail. The
influence of tensile bending stresses at the underhead radius (UHR), which are
associated with contact loads along with a combination of the lateral bending of the
whole rail profile and local bending of the head-on-web are included. This effect is of
greater importance under high axle load conditions and for increased head wear, as was
demonstrated in chapter 5. Furthermore bending of the rail head-on-web due to the
eccentricity of contact load was not considered in the crack growth investigations
conducted by Fletcher et al [27, 119] and Dang Van et al [28]. Jeong et al [6] ,
however, did include this behaviour in an examination of the growth behaviour of
reverse detail fractures that were initiated at the lower gauge corner of heavily-worn
rail. Relatively little work has been done on long cracks that approach the critical crack
length for rapid fracture, as reported by Kapoor et al [23]. In the investigation
conducted for this thesis, a long crack was introduced at the gauge corner. The unstable
growth behaviour of the crack under various situations of contact load applied to the rail
Chapter 7. Rapid fracture modelling using X-FEM
149
head at the varied eccentric locations was evaluated with respect to different rail worn
profiles.
7.2 Crack model development
A detailed literature survey revealed possible methods for an analysis of stress intensity
factors that included but was not limited to finite element modelling (FEM) [114, 150,
for the modelling of cracks whose geometry is independent of the original finite element
region [177].
Heaviside function H(x) is used to represent a displacement jump across the crack face
and is given by the following expression:
( ) { ( )
} (7.1)
Where
is a sample (Gauss) point,
is the point on the crack closest to , and
n is the unit outward normal to the crack at , as shown in Figure 7.1 [135].
Figure 7.1 Illustration of normal and tangential coordinates for a smooth crack [135]
Discontinuous geometry using the level set method 7.2.2
In addition, the displacement field approximation was divided into continuous and
discontinuous parts. The Level Set Method (LSM) was used to locate the discontinuity
by defining the crack geometry by two orthogonal signed distance functions, as given
below:
Chapter 7. Rapid fracture modelling using X-FEM
151
Ļ describes the crack surface and
Ļ is used to construct an orthogonal surface, as shown in Figure 7.2.
The intersection of the two surfaces gives the crack front.
n+ indicates the positive normal to the crack surface, m+ indicates the positive normal
to the crack front, as shown in Figure 7.2 [135].
Figure 7.2 Representation of a non-planar crack in three dimensions by two signed distance functions Ļ and Ļ [135]
X-FEM crack model 7.2.3
The rapid fracture behaviour was examined using several new models by inserting X-
FEM elements into the existing finite element (FE) meshes, which were previously
validated by comparison with in-track measurements, as given in chapters 3 and 5. For
this thesis, extended finite element method (X-FEM) elements [135] were applied to
model the crack. The X-FEM elements were applied to the crack region, which was
directly located above the underhead radius (UHR) reference location, as shown in
Figure 7.3a-d; the remainder of the model used FEM elements. This newly developed
numerical model utilized the commercial code ABAQUS 6.11-2 to examine the
Chapter 7. Rapid fracture modelling using X-FEM
152
unstable crack growth behaviour. The ABAQUS 6.11-2 X-FEM models solutions were
conducted on 2.99 GHz CPU.
The extended finite element method X-FEM in ABAQUS is a mesh independent
method which is based on the enriched elements and its capabilities were extensively
verified in the literature [177]. The critical crack sizes were considered with dense mesh
as shown in Figure 7.3c-d. The validation of crack modelling using track data and the
comparison of the crack sizes using the operational fractures was not possible in the
absence of field data.
The X-FEM model used a single 68 kg/m rail with 6 m length laid on discrete elastic
foundations to parametrically study the unstable growth behaviour of RCF cracks
subjected to change of the head wear (HW), the contact patch offset (CPO) from the rail
centreline, and the (L/V) ratio of lateral (L) to vertical (V) loads.
The X-FEM model considered a single crack, which was in the transverse orientation
with an approximate shape based on the typical transversally deviated head checks that
occur under heavy haul conditions (Figure 1.2). The crack face friction is not included
in the current model. The crack shape is described by the semicircular crack radius R,
and the crack surface length 2R (Figure. 7.3a). The crack has a length of R with an
angle to the top surface of rail as in Figure 7.3b and is centered at an offset of 27 mm
from the centerline of the rail head cross-section, as shown in Figure 7.3c. The change
in angle Īø was also investigated using the simulations described in the next paragraph.
A single wheel load was simulated, applying a fully slipping Hertzian contact patch on
the wheel-rail contact surface. The contact loads included lateral and vertical loads (L/V
ratio of lateral to vertical load), a contact patch offset (CPO) from the rail centerline
(Figure 7.3c) and contact patch movement from the crack location along the rail
longitudinal running direction (Figure 7.4). A discrete elastic foundation directly at the
location of the sleepers was used to support the rail. The details of the model parameters
were given in Table 7.1. The finite element modeling details have already been
discussed in chapters 3 and 5. It should also be noted that the details on input
Chapter 7. Rapid fracture modelling using X-FEM
153
parameters such as loading, boundary conditions and support characteristics will remain
the same as was discussed in chapter 3.
Figure 7.3 A single crack at the gauge corner: (a) crack shape (crack length = R and surface length = 2R); (b) crack orientation ( = 70o); (c) fine mesh at highlighted enriched region including crack and contact patch positions; (d) Contact patch offset (CPO), L/V ratio and position of the crack from the centre of rail head cross-section
(a)
(c) (d)
(b)
Contact patch positions
Crack
Contact patch
Crack front
Crack
Chapter 7. Rapid fracture modelling using X-FEM
154
Figure 7.4 Loading steps at different positions relative to the crack
The simulations incorporate several steps to enable the examination of unstable crack
growth behaviour for different loading positions relative to the position of the crack
(Figure 7.4). The passing wheel was simulated by applying the contact loading at step 1,
then at step 2, then step 3, and so on. The stress intensity factor (SIF) in the plane
opening Mode I (KI) was calculated along the crack front for each of the locations. For
each assessment, the size of the crack and the rail head wear (HW) were defined as two
variables in which the crack length, R was in the range of 5 mm to 30 mm, and HW was
in the range of 10 mm to 22 mm, respectively, and the crack orientation ( Īø ) was 60o,
70o and 90o. The results from the parametric analyses were used to evaluate unstable
crack propagation in the rail.
Chapter 7. Rapid fracture modelling using X-FEM
155
Table 7.1 Crack model parameter values
Model Parameters Value Rail span, S (mm) 6000 Axle load, V (kN) 343.4 L/V ratio 0, 0.2 Worn rail profile / head wear, HW (mm) 10, 15, 22 Crack radius, R 22, 17, 10 Crack orientation / inclination, Īø (o) 60o, 70o and 90o Contact patch offset, CPO (mm) 0,15, 20 Semi major axis, a (mm) 10 Semi minor axis, b (mm) 4 Youngās modulus, E (MPa) 209,000 Poissonās ratio, 0.3
7.3 Variations of KI max during loading cycle
The position of the crack front was normalized by the total length of the crack front.
The crack front was positioned from 0 at the underhead radius (UHR) of the lower
gauge corner point to 1 at the point near the top surface of the rail (Figure 7.3a). The
stress intensity factors (SIF) (KI, KII and KIII) at the crack tip were used to examine the
rapid fracture behaviour in response to the stress distribution resulting from the
combination of contact loads and bending stresses. Three commonly used rail grades: a
plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat Treated (LAHT) grade
[127], and a Hypereutectoid (HE) heat treated grade [81], were used to examine the
effect of material properties on the rapid fracture of the crack.
Effect of contact patch offset (CPO) and L/V ratio 7.3.1
Two representative conditions can be used to simulate the effect of vehicle curving and
hunting: the contact patch offset and L/V ratio. The contact patch lateral displacement
was considered in this study for two reasons. Firstly, the vehicle tends to exhibit lateral
Chapter 7. Rapid fracture modelling using X-FEM
156
movement due to steering and hunting behaviour and, secondly, the contact patch offset
may change because of a change in the position of running band due to a reduction in
the cross-section area because of natural wear or after profile grinding. In order to study
the effect of contact patch location, an L/V ratio of 0 was taken to minimise the effect of
the lateral load (L). Figure 7.5 shows the maximum (over the entire crack front) Mode I
stress intensity factor (KI max) under contact loads of L/V of 0, 0.2 when applied to
different loading positions relative to the position of crack length R = 17 mm, = 90o
and HW 15 mm.
The fracture region is defined by KI max > KIC. The three fracture toughness (KIC) values
mentioned for a plain C-Mn Head Hardened grade, HH - KIC = 45 MPam [80], a Low
Alloy Heat Treated grade, LAHT3 - KIC = 39.1 MPam [127], and a Hypereutectoid
heat treated grade, HE - KIC = 35.4 MPam [81] rail steels are also illustrated in Figure
7.5 by the three dashed lines, and are used to examine the effect of material properties
on rapid fracture of the crack. When the contact load is directly imposed above the long
crack (R = 17 mm), for CPO =15 mm, L/V = 0, KI max value is small, below the fracture
region defined by KI max > KIC , as shown in Figure 7.5. This indicates that rail failure is
unlikely under these loading and rail head wear conditions.
The results shown in Figure 7.5 reveal that rapid fracture could happen when the contact
load is directly imposed above the long crack (R = 17 mm) for CPO = 20 mm, L/V =
0.2, as the KI max > KIC (KIC = 35.4 MPam for HE rails, KIC = 39.1 MPam for LAHT3
rails, and KIC = 45 MPam for HH rails). This can result in the rapid fracture of rail
under these loading and head wear conditions. The crack is driven by local tensile
bending stress having a spike at the underhead radius. The loads applied near to the rail
gauge corner (CPO = 20 mm) provide higher tensile bending stress than the loading
case of L/V = 0 at a contact patch offset (CPO) of 15 mm. As a result, KI max increases
up to 71 MPam, being almost three times higher than the case with CPO 0 mm. When
the loads are positioned away from the crack, KI max reduces significantly, as the local
bending stresses are absent at the crack front and the global bending stresses are small,
as shown in Figure 7.5.
Chapter 7. Rapid fracture modelling using X-FEM
157
Figure 7.5 The maximum KI (KI max) at the crack front (R = 17 mm, = 90o) under the different loading positions
Effect of crack length on SIFs 7.3.2
Under the same conditions (CPO = 20 mm, L/V = 0.2, = 90o and HW = 15 mm), the
crack length, R, is a key factor influencing stress intensity factors at the crack front, as
KI max value is dependent on the crack length R. Figure 7.6 shows that the crack with a
length R = 17 mm for HW of 15 mm, produces the highest value of KI max compared
with a crack length R = 5 mm and 30 mm. This is because the crack front for R = 17
mm crack is located at the underhead radius where the tensile stress is much higher,
whereas the relatively smaller crack (R = 5 mm) is away from this highly stressed
region. Under the extreme loading conditions (L/V = 0.2 and CPO = 20 mm), the crack
may not extend to 30 mm, as the probability is high for a rapid fracture at R = 17 mm.
Therefore, a long crack of 30 mm may be inappropriate for use as it is away from the
underhead radius and the possibility of rapid failure occurs only at the 17 mm long
The model used a wheel load of 172 kN (Table 7.1), which is equivalent to an axle load
of 35 tonnes. All the other influencing parameters such as boundary conditions and
support characteristics values are same as was given in Chapter 3. It focused on the
effect of HW conditions; contact patch offset (CPO) and L/V representative of curved
track. The contact patch offset (CPO) and L/V ratio are two representative conditions
that were used to simulate the effect of vehicle curving and hunting. Further, the CPO
changes due to the offset of running bands as the rail HW profile changes because of a
reduction in the cross-sectional area, either due to natural wear or after profile grinding.
Under these conditions the stress range in the underhead radius is higher for the rail of a
curved track due to the lateral transverse bending and torsion of the head-on-web.
Additional sets of finite element analyses referring to contact patch offset
(representative of lateral wheel positions), L/V ratios (representative of rail curvature)
and head wear (HW) conditions were carried out as described in the next section. The
crack growth behaviour, especially the probability of rapid (unstable) crack growth
behaviour, was examined for three rail grades commonly used in heavy haul
applications: a plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat Treated
(LAHT) grade [127], and Hypereutectoid (HE) Heat Treated grade [81].
Chapter 8. RCF cracks under mixed-mode loading
173
Mixed-mode fracture criteria 8.2.1
The literature suggests a number of fracture criteria to use when investigating how a
pre-existing mixed-mode crack becomes unstable. For example, those introduced by
Erdogan and Sih [186], Schollmann et al [187], Hussain et al [188] and Richard et al
[189-191] are well-known criteria for analysing a fracture. Richardās criterion has
shown to be numerically accurate and also simple and practical to apply. It is based on
approximate formulae and has proven to be in good agreement with experiments and
other accurate criteria such as the Schollmann maximum principal stress criterion [187,
191, 192].
Figure 8.3 Development of fracture surface for Mode I, Mode II, Mode III and mixed-mode-loading of cracks [189, 190]
The stress intensity factors (SIFs) in Mode I, II and III (KI, KII and KIII respectively)
were calculated along the crack front for each of the contact loading locations, as shown
in Figure 8.2a. The equivalent mixed-mode SIF (Keq ) (Equation 8.1) due to the
variations of the individual stress intensity factors can be found from Richard et al [189-
191], as shown in Figure 8.3.
Chapter 8. RCF cracks under mixed-mode loading
174
ā
( ) ( ) (8.1)
According to the criterion proposed by Richard et al [189, 190], the unstable fracture
occurs if is equal to the fracture toughness KIC of the material. i.e.
= KIC (8.2)
8.3 Variations of SIFs (K I, K II, K III) during loading cycle
A single long crack formed at the gauge corner of the rail was analysed first. The crack
was in the transverse orientation of angle with a semi-circular shape approximating a
typical transversally deviated head check, such as the one shown in Figure 8.1b, with a
crack radius R, and a crack surface length of 2R (see Figure 8.2a). The position on the
crack front was normalized by the total length of the crack front from the position 0
corresponding to the underhead radius near the lower gauge corner point to the position
1 at a point of the top surface of the rail. The passage of the wheel was considered using
the longitudinal movement of the associated contact patch and pressure, as shown in
Figure 8.2a. The resultant SIFs (KI, KII , KIII) at the crack front were used to examine the
unstable growth behaviour of this crack in response to the local bending stress
distribution resulting from the mixed-mode loading conditions of the contact and
bending stresses.
Figures 8.4 (a)-(d) show the variation of stress intensity factors (maximum over the
entire crack front) at each longitudinal position of the wheel over the rail running
surface. Figures 8.4(a) and (c) show this variation for a contact patch offset of 0,
whereas Figures 8.4 (b) and (d) are for a value of 15 mm. The maximum value was
observed when the wheel passed with a CPO of 15 mm (Figure 8.4b). Figure 8.4b
shows the maximum stress intensity factors in the tensile opening (KI), and the in-plane
sliding (KII) modes are much higher than the out-of-plane tearing (KIII) mode. The
maximum stress intensity factors in the two in-plane modes correspond to the position
of the wheel close to the top of the crack, as shown by points A and B in Figure 8.4b.
Chapter 8. RCF cracks under mixed-mode loading
175
The plots in Figures 8.4 c and d show these small changes in values in detail. The
maximum in-plane opening mode KI max (point A in Figure 8.4 d) corresponds to the
position when the wheel is over the crack, and is the result of tensile stresses due to the
lateral bending of the whole rail profile and the localised vertical and lateral bending of
the head-on-web, as shown in Figure 8.4b and d. When the loads are positioned away
from the crack in the longitudinal direction, the KI max reduces as the local bending
stresses are absent at the crack front and the global bending stress is small.
(a) CPO 0 mm (b) CPO 15 mm
(c) CPO 0 mm (d) CPO 15 mm
Figure 8.4 Variation of maximum (over the entire crack front position) SIFs during one wheel passage (loading cycle) with different contact patch offsets (CPO), L/V = 0
For a wheel passing on the rail head without eccentricity (CPO = 0 mm), as shown in
Figures 8.4 (a) and (c), these changes in values are small, due to the absence of local
bending stresses. As with the contact patch offset CPO = 0, the head is in compression
due to global bending. KI is negative and shows compressive values, as depicted in
Figure 8.4a. The changes in values due to global bending stresses are lower compared to
local bending and are almost similar in both the loading cases, as shown in Figures 8.4
(a) and (b).
8.4 Variations of Keq along crack front during loading cycle
An equivalent SIF (Keq) is used to predict the fracture behaviour of RCF cracks by using
the Equation 8.1. The equivalent (Keq) SIF along the crack front changes, due to the
changes in the values of individual stress intensity factors when a wheel passes the
crack. The wheel position relative to the position of the crack is denoted by the index
X/a and is used in the subsequent investigation. X is the distance from the contact
pressure peak to the crack mouth and āaā is the semi-major axis of the contact patch, as
shown in Figure 8.2d and the value of āaā is stated in Table 7.1.
Effect of contact patch offset (CPO) 8.4.1
In order to study the effect of CPO, a constant L/V ratio of 0 was taken to minimise the
effect of the lateral load (L). For the first case to be described, Figure 8.5 shows the
SIFs at a crack tip when a wheel is passing on the rail head without eccentricity (CPO =
0 mm). No lateral traction (L/V = 0) is included. The crack (R = 10 mm, inclined at =
90o) to the rail running surface with rail HW of 22 mm was considered. The stress
intensity factors (KI , KII , KIII, Keq) are shown in Figure 8.5 at the normalised crack front
position during a wheel passage. Due to global bending at the contact patch offset CPO
= 0, for X/a = 0, the head is seen in compression. KI was negative and showed
compressive values, as depicted in Figure 8.5a and these negative KI values are treated
as zero for Keq calculations. Figure 8.5 d shows that Keq values are very small, far below
the fracture region defined by Keq KIC.
Chapter 8. RCF cracks under mixed-mode loading
177
Figure 8.5 SIFs and Keq at the crack front when a wheel passes with CPO = 0 mm, L/V = 0 for crack (R = 10 mm, = 90o) to the rail running surface with HW 22 mm
Figure 8.6 depicts the second case, when a wheel with the CPO 15 mm, L/V =0, passes
the crack (R = 17 mm, = 90o) on the rail surface with HW 15 mm. According to the
SIF values, a significant variation at normalised crack front positions is seen compared
to the case with CPO = 0 mm (Figure 8.5). Figure 8.6a depicts the SIF due to tensile
opening (KI) behaviour at the crack front. When the wheel is directly on top of the crack
mouth (X/a = 0), the maximum value of SIF in Mode I (KI max) is at the 0 crack front
position corresponding to the underhead radius near the lower gauge corner point. Other
wheel positions show a lower value of KI as compared to X/a = 0. The Mode II stress
intensity factor (KII) is plotted in Figure 8.6b. When the wheel approaches the crack, the
maximum value of SIF in sliding (KII max) is positive at the 0 crack front position (X/a
=1.5) and starts decreasing and becomes almost 0 when the wheel is over the crack (X/a
0
5
10
15
20
25
30
35
40
45
50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Keq
(MP
aām
)
Normalised crack front position
X/a = 1.5
X/a = 0.6
X/a = 0
X/a = -1
X/a = -1.7
HH - KIC
LAHT - KIC
HE3 - KIC
-5
0
5
10
15
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-100
-80
-60
-40
-20
0
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KI(M
Paā
m)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-10
-5
0
5
10
15
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
I(M
Paā
m)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
Fracture Region
(b)
(c)
(a)
(d)
Chapter 8. RCF cracks under mixed-mode loading
178
=0). When the wheel passes the crack the KII increases, but with a negative sign, as
depicted in Figure 8.6b.
Figure 8.6c shows the values of the SIFs in the out of plane tearing mode (KIII), which
are much lower compared to the opening (KI) and the sliding (KII) modes, suggesting
that the in-plane opening and shear forces play an important role in the unstable
propagation of the crack at the underhead radius.
Figure 8.6 SIFs and Keq at the crack front when a wheel passes with CPO = 15 mm, L/V = 0 for crack (R = 17 mm, = 90o) to the rail running surface with HW 15 mm
Figure 8.6d suggests that the maximum Keq occurs at the loading step (X/a = 0), when
the loading is imposed directly above the long crack, as shown in Figures 8.2 (a) and
(d). Equation 8.1 shows that Keq is primarily dependent on the value of KI. Figures 8.6
(a) and (d) show that as the KI value rises, so does the Keq value. This is due to the
0
10
20
30
40
50
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
K eq
(MPa
ām)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
HH - KIC
LAHT3- KIC
HE3 - KIC
-40
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
(MPa
ām)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KI(M
Paām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
I(M
Paām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0 X/a = -1X/a = -1.7
(a)
Fracture Region
(c)
(b)
(d)
Chapter 8. RCF cracks under mixed-mode loading
179
bending of the rail head-on-web forced by the eccentricity of loads from the rail
centreline. As a result of changes in the stress state and this tensile stress, KI increases
and so does Keq. Figure 8.6d shows that Keq values are below the fracture region defined
by Keq KIC, which indicates that the risk of rail failure is unlikely under these loading
and rail head wear conditions.
The third case examined is when a wheel with CPO = 15 mm, L/V =0, passes the crack
(R = 10 mm, inclined at = 90o) on the rail surface with HW 22 mm, as shown in
Figure 8.7. The higher head wear leads to a higher in-plane opening SIF (KI) at 0 crack
front position and therefore a higher Keq. The in-plane opening is attributed to higher
tensile stress as a result of an increase in the head wear, when the loading is applied at a
contact patch offset from the centre of the rail head cross-section. Moving the contact
location towards the gauge corner, away from the centre of the rail head cross-section,
leads to an increase in the tensile stress at the underhead radius. The combination of
increased bending stress due to increased localised bending of the rail head-on-web and
lateral bending of the head opens the crack (positive KI ) when the wheel contact
pressure peak is directly above the crack mouth (X/a = 0) and is exacerbated by
increasing head wear. This reveals that the in-plane opening SIF (KI) will be the
dominant driver of fracture. This is due to a linear increase in the in-plane tensile
opening SIF (KI) at the underhead radius with the highest value at the 0 crack front
position, as shown in Figure 8.7. Similarly the Keqmax is highest at the same location (see
Figure 8.7d) due to the large contribution of the positive KI mode.
The fracture region is defined by Keqmax > KIC. Fracture toughness (KIC) values of three
high strength rail grades were used in this analysis. The fracture toughness (KIC) value
used for a plain C-Mn Head Hardened grade [80], HH - KIC = 45 MPam is reported by
Marich et al [179] and Skyttebol et al [151]. For a Low Alloy Heat Treated grade [127],
LAHT3 - KIC = 39.1 MPam, and Hypereutectoid heat treated grade [81], HE - KIC =
35.4 MPam; these values are estimated by Ueda et al [130]. The representation of
these values is also illustrated in the Figure 8.7d by three dashed lines, which are used
to examine the effect of material properties on the rapid fracture of the crack. The effect
of increasing the fracture toughness of the rail material for the same range of loading
Chapter 8. RCF cracks under mixed-mode loading
180
conditions is also shown in Figure 8.7d. Increasing the fracture toughness enables the
limit to be increased for high strength rail material grades. These values are all based on
previous studies (as reported in the respective references), and are used solely to
illustrate the effect of differences due to rail grade. A more rigorous analysis would
require testing of the rail material grades under identical test conditions, to ensure that
any differences in KIC were not attributable to the test methods used.
Figure 8.7 SIFs and Keq at the crack front when a wheel passes with CPO = 15 mm, L/V = 0 for crack (R = 10 mm, = 90o) to the rail running surface with HW 22 mm, for three rail grades: a plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat Treated (LAHT) grade [127], and Hypereutectoid (HE) Heat Treated grade [81]
0
10
20
30
40
50
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Keq
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
HH - KIC
LAHT3- KIC
HE3 - KIC
-40
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-40
-20
0
20
40
60
80
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KI(M
Paā
m)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-40
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
I(M
Paā
m)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0 X/a = -1X/a = -1.7
Fracture Region Increasing the fracture strength
(a)
(c)
(b)
(d)
Chapter 8. RCF cracks under mixed-mode loading
181
A comparison of equivalent SIF (Keq) with fracture toughness (KIC) values (i.e. Keq max >
KIC , where KIC = 35.4 MPam for HE rails, KIC = 39.1 MPam for LAHT3 rails, and
KIC = 45 MPam for HH rails) indicated that a rapid fracture is likely to occur for these
rail material and loading conditions.
Effect of L/V ratio 8.4.2
The next case to be described is when the wheel passes the crack (R = 17 mm, inclined
at = 90o) with the lateral load (L) due to the angle of attack between the wheel and the
rail. A lateral load is generally influenced by both the vehicle steering effects on the
curved track and the hunting behaviour of the vehicle due to the vehicle dynamics, and
it can be scaled by the L/V ratio. An L/V ratio of 0 represents loading conditions in
tangent track, however this may increase slightly in the presence of vehicle hunting and
a value of 0.1 or even higher may be considered. .An L/V ratio of 0.2 represents a
loading condition in a curved track, as was reported by Marich [36, 37]. The high rail
experiences outward lateral shear traction and could reach a value of 0.4. The value of
L/V = 0.2 used for this thesis is on the lower side and simulates the inwards lateral shear
traction, rather than the outwards lateral shear traction considered in the work
referenced in [38, 42-45].
The lateral load can change the stress intensity factor pattern significantly at the
transverse rail curvature, as shown in Figure 8.8, with an L/V ratio = 0.2 applied at a
contact patch offset (CPO) of 15 mm for a crack (R = 17 mm, angle = 90o) to the rail
running surface with HW of 15 mm. The in-plane opening (KI) and sliding (KII) SIF are
plotted in Figures 8.8a and 8.8b. An increase in L/V = 0.2 leads to higher KII and the
pattern of KII is symmetric but with higher values, as expected. When the wheel is
directly on the top of the crack, the expected pattern of KI also has higher values,
particularly at the 0 crack front position. This higher tensile bending stress at the
underhead radius is caused by the superposition of eccentric loading with lateral shear
traction, as was expected and is shown in Figure 8.8a. The variation of SIFs in the out-
of-plane tearing mode KIII is similar to the last case, as shown in Figure 8.8c, presenting
much lower values when compared to KI and KII, even if the lateral traction is included.
Chapter 8. RCF cracks under mixed-mode loading
182
Figure 8.8d depicts a detailed analysis regarding the fracture behaviour of the rail
underhead radius. Increasing the L/V ratio from 0 to 0.2 with contact patch eccentricity
(CPO = 15) shows that on the lower corner of the gauge face of the rail, the
combination of tensile bending stress due to superposition of eccentric and inward
lateral loading opens the crack (positive KI ) when the wheel is directly on the top of the
crack (X/a = 0). Similarly, the Keq is highest at the same location. When the Keqmax > KIC
for all the three rail material grades is considered, as shown in Figure 8.8d, the rail will
break with a rapid fracture.
Figure 8.8 SIF at the crack front when a wheel passes with CPO = 15 mm, L/V = 0.2 for crack (R = 17 mm, = 90o) to the rail running surface with HW 15 mm
0
10
20
30
40
50
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Keq
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
HH - KIC
LAHT3- KIC
HE3 - KIC
-40
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-20
0
20
40
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KI(M
Paā
m)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
I(M
Paā
m)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0 X/a = -1X/a = -1.7
Fracture Region Increasing the fracture strength
(a)
(c)
(b)
(d)
Chapter 8. RCF cracks under mixed-mode loading
183
Results of extreme loading conditions (HW = 22 mm, CPO = 20 mm, and L/V =0.2) on
a heavily worn rail head are shown in Figure 8.9. As with the last case, a higher L/V
=0.2 leads to a higher Keq and is strongly correlated with the lateral load, which results
in a higher tendency to rapid fracture leading to catastrophic rail failure.
Figure 8.9 SIF at the crack front when a wheel passes with CPO = 20 mm, L/V = 0.2 for crack (R = 10 mm, = 90o) to the rail running surface with HW 22 mm
This modelling showed how an inwards lateral shear traction of L/V = 0.2 can expedite
the rapid fracture of a pre-existing crack on a heavily worn rail of HW 15, 22 mm with
CPO of 15, 20 mm, as shown in Figures 8.8 and 8.9 respectively. This investigation
revealed that the lateral load (L/V) with a contact patch offset (CPO) could play a
significant role in the rapid fracture behaviour of pre-existing RCF cracks extending to
the rail underhead radius, if a heavily worn rail exists in practice. In the field, most of
0
20
40
60
80
100
120
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Keq
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
HH - KIC
HE3 - KIC
LAHT3- KIC
-40
-20
0
20
40
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
(MPa
ām)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-100
-80
-60
-40
-20
0
20
40
60
80
100
120
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KI
(MPa
ām)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-40
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
I(M
Paām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0 X/a = -1X/a = -1.7
Fracture Region
(a)
(c)
(b)
(d)
Chapter 8. RCF cracks under mixed-mode loading
184
the cracks occurred on the high rail in curves of decreasing radii, around 850m ā 350m,
as was reported by Kerr et al [193], and these are more prone to rail fracture. The
curved rails are heavily worn out, with contact patch eccentricity due to steering effects
and cant deficiency as compared to tangent tracks. If these curves are poorly lubricated
then lateral shear traction may rise and contribute to rail failure due to rapid fracture.
Comparison of head wear (HW) 8.4.3
The value of Keq along the crack front, when a wheel with different loads passes a crack
( = 90o to the rail running surface) with different head wear, is utilized to predict the
safe head wear limit. Figures 8.10 (a)-(b) show the variation in Keq with increasing head
wear and a range of crack sizes (R) and loading conditions. It should be noted from
Figure 8.10a that the least head wear (HW =10 mm) presents the lowest Keq for the
same loading condition due to the small values of tensile stress seen at the underhead
radius. At the underhead radius from the normalized crack front position of 0 to 0.2, Keq
is positive but Keq KIC for the HH, HE3 and LAHT3 rail grades considered in this
thesis, so no rapid fracture was expected.
Figure 8.10a also shows the stress intensity factors for a crack with heavily worn rail
profiles (HW 15 and 22 mm). In these cases, Keq exceeds the fracture toughness of the
HH, HE3 and LAHT3 grades from the normalized crack front position of 0 to 0.153 at
the underhead radius. Hence, rapid (unstable) crack propagation may develop for these
combinations of the crack length and head wear.
A rapid fracture can be predicted with 15 mm of head wear at an L/V of 0.2 and contact
patch offset of 20 mm (Figure 8.10a). The corresponding HW limit for an L/V ratio of 0
and contact patch offset of 15mm is approximately 22 mm, as shown in Figure 8.10b.
The safe HW limits are also summarised in Table 8.2. The prediction of safe HW limits
was based on the three HW profiles considered. Further study is required with multiple
sets of analysis with a range of HW, crack size, orientations, and loading conditions, in
order to examine the sensitivity of these limits in detail.
Chapter 8. RCF cracks under mixed-mode loading
185
(a) CPO 20 mm, L/V = 0.2
(b) CPO 15 mm, L/V = 0
Figure 8.10 SIF at the crack front when a wheel passes for crack angle = 90o to the rail running surface with different HW, crack sizes R, CPO and L/V, for three rail grades: a plain C-Mn Head Hardened (HH) grade [80], a Low Alloy Heat Treated (LAHT) grade [127], and Hypereutectoid (HE) Heat Treated grade [81]
0
20
40
60
80
100
120
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7K
eq(M
Paā
m)
Normalised crack front position
HW = 22 mm, R = 10 mmHW = 15 mm, R = 18 mmHW = 10 mm, R = 23 mm
HE3 - KIC
HH - KICLAHT3 - KIC
0
10
20
30
40
50
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
K eq
(MPa
ām)
Normalised crack front position
HW = 22 mm, R = 10 mmHW = 15 mm, R = 17 mm
HE3 - KIC
HH - KIC
LAHT3 - KIC
R = 22 mm
R = 10 mm R = 17 mm
Fracture Region Increasing the fracture strength
Increasing the fracture strength
No rapid fracture
Rapid fracture (d)
No rapid fracture R = 10 mm R = 17 mm
Fracture Region
Chapter 8. RCF cracks under mixed-mode loading
186
Table 8.2 Predicted head wear limits based on fracture strength with extreme loading cases representative of different track conditions [valid for studied crack size (radius) R =10 mm, 17 mm and 22mm as shown in Figures 8.2 and 8.10]
Rail track curvature Assumed L/V
ratio Patch offset (mm) Safe HW limit (mm)
(predicted)
Tangent (prone to hunting) / shallow radius 0 15 15
Tight radius curve 0.2 20 10
Crack shielding by multiple cracks 8.4.4
The effect of crack shielding by multiple cracks was examined. This was done by
considering clusters of both 3 and 7 cracks with a spacing s = 5 mm from the middle
crack, as shown in Figure 8.2 (a)-(c). This spacing was chosen as representative of the
separation of cracks seen on curved rails prone to RCF cracks [27]. The SIF predicted at
the crack front for the central crack of seven multiple cracks when a wheel passes with
CPO = 15 mm, L/V = 0 for crack angle = 90o to the rail running surface with HW of
22 mm is shown in Figure 8.11.
Figure 8.7 presents the results for the SIF for a single crack under a CPO 15 mm, with
an L/V = 0 and head wear of 22mm. The results for 7 cracks are shown in Figure 8.11a.
A comparison of these results suggests that there is a reduction in the maximum KI of
the middle crack of around 37% relative to the single crack. The results in Figure 8.11b
indicate that the maximum KII is reduced by 19 % of the single-crack value (Figure
8.7b). The KIII values for single and seven crack cluster models were close. The results
in Figure 8.11d indicate that the maximum Keq is reduced by 19 % from the single-crack
value (Figure 8.7d). The maximum of Keq was found to be sensitive to the presence of
multiple cracks to a similar degree to the maximum of KI and KII values, as shown in
Figure 8.11d.
Chapter 8. RCF cracks under mixed-mode loading
187
Figure 8.11 The effect of modelling seven multiple cracks on SIF predicted for the central crack with SIF on crack front when a wheel passes with CPO = 15 mm, L/V = 0 for crack angle 90o to the rail running surface with HW 22 mm
The Keq for a single crack and clusters of 3 and 7 cracks were compared. The Keq
predicted on a normalised crack front for the central crack, as used to model 1, 3 and 7
multiple cracks, is shown in Figure 8.12. The results for the reduction in the maximum
of Keq for normalised crack front position of 0 are taken for comparison. The results for
L/V = 0 (Figure 8.12a) suggest that when compared to a single crack, there is a 27%
reduction in the Keq max of the middle crack when the model with 3 cracks is considered.
Similarly the reduction in Keq max of the middle crack is around 35% when 7 cracks are
considered in comparison to the single crack model. In the presence of these reductions
in SIFs, it can be seen that the multiple cracks shield each other and prevent the rail
fracturing through a reduction in the stress intensity factors. For the 3-crack model, the
predicted Keq max just reached the KIC for HE and LAHT3 rail grades, for a head wear of
0
10
20
30
40
50
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Keq
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
HH - KIC
LAHT3- KICHE3 - KIC
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
(MP
aām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-40
-20
0
20
40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KI(M
Paām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0X/a = -1X/a = -1.7
-20
-10
0
10
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
KII
I(M
Paām
)
Normalised crack front position
X/a = 1.5X/a = 0.6X/a = 0 X/a = -1X/a = -1.7
Fracture Region Increasing the fracture strength
(a)
(c)
(b)
(d)
Chapter 8. RCF cracks under mixed-mode loading
188
22 mm. Nevertheless, the 7 crack model shows Keq max < KIC for both LAHT3 and HH
rail grades, which is a significant reduction when compared to the single crack case.
(a) L/V = 0 (b) L/V =0.2
Figure 8.12 The effect of modeling multiple cracks on Keq SIF predicted for the central crack when a wheel passes with CPO = 15 mm, for crack angle 90o to the rail running surface with HW 22 mm; L/V 0, 0.2.
The results for L/V = 0.2 (Figure 8.12b) suggest a subsequent increase in the Keq max due
to increased tensile stresses at the underhead radius, and as a result of the superposition
of the load eccentricity with lateral shear traction, as shown in Figure 8.12b. Compared
to a single crack, there is a 43% reduction in Keq max of the middle crack if the model
with three cracks is considered. Similarly, the reduction in Keq max of the middle crack is
around 50 % if seven cracks are considered, as compared to the single crack model.
In this doctoral study, multiple cracks of the same size and shape were modelled, which
is not what is generally observed in the field. It has been commonly observed that one
crack in the series can become longer compared to others. In this way, there is a
possibility that the shielding effect may be reduced and a single crack may become
critical, resulting in the rapid fracture of the rail, as presented in the case of single crack
models. Therefore, the single crack case is more critical and the analysis and design
against failure of a single crack should be considered accordingly.
0
10
20
30
40
50
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
K eq
(MPa
ām)
Normalised crack front position
Single crack3 cracks7 cracks
HH - KIC
LAHT3- KIC
HE3 - KIC
0
10
20
30
40
50
60
70
80
90
100
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7K e
q(M
Paām
)
Normalised crack front position
Single crack3 cracks7 cracks
HH - KIC
LAHT3- KICHE3 - KIC
Fracture Region Fracture Region Increasing the fracture strength
Chapter 8. RCF cracks under mixed-mode loading
189
8.5 Summary
This chapter has discussed the prediction of the rapid fracture behaviour of rails
associated with local bending of the rail head-on-web. A method has been developed to
predict unstable growth of multiple RCF cracks under multi-axial loading at the gauge
corner of the rail. The extended finite element method (X-FEM) model was used to
investigate the stress intensity factor near the crack front that results from contact and
bending stresses. The equivalent stress intensity factor was considered in order to
investigate the fracture behaviour. Both single and multiple cracks were modelled at
different lengths and orientations, over a range of head wear conditions.
The results of the rapid fracture analysis using mixed-mode loading suggest that when a
wheel passing on the rail head without eccentricity, and approaches a position above the
crack mouth, the head is in compression as a result of global bending. In-plane tensile
opening mode, KI is negative and the Keq values are very small, far below the fracture
region defined by Keq KIC. Moving the contact patch towards the gauge corner away
from the centre of the rail head cross-section increases the tensile stress at the underhead
radius. This is due to bending of the rail head-on-web forced by the eccentricity of loads
from the rail centreline. The maximum of Mode I stress intensity factor KI max occurs
when the contact loading is applied directly above the crack location. As a result of
changes in the stress state, the KI increases and so does the Keq.
Higher head wear leads to a higher in-plane opening stress intensity factor (KI) at the
crack front position of 0, corresponding to the underhead radius near the lower gauge
corner point, and it therefore has a higher Keq. The in-plane tensile opening behaviour
can be attributed to higher tensile stress as a result of an increase in the head wear, if the
loading is applied at the contact patch offset from the centre of the rail head cross-
section. However the L/V ratio also has a significant effect on KI and KII and
consequently Keq. This is due to the higher tensile bending stress at the underhead radius
caused by the superposition of the eccentric loading with inwards lateral shear traction,
as was expected. The values of the stress intensity factors in the out-of-plane tearing KIII
mode are much lower, as compared to the opening KI and the sliding KII modes,
Chapter 8. RCF cracks under mixed-mode loading
190
suggesting that the in-plane opening and shear forces play an important role in the
unstable propagation of pre-existing RCF cracks that extend to the rail underhead
radius. Therefore, KI max and KII max were more significant in terms of rapid fracture than
KIII max.
The results revealed that the lateral load (L/V) with a contact patch offset (CPO) could
play a significant role in the rapid fracture behaviour of pre-existing RCF cracks
extending to the rail underhead radius, if a heavily worn rail exists in practice.
Therefore, pre-existing RCF cracks at the gauge corner region with crack lengths
extending to the underhead radius region could increase the risk of rapid (unstable)
fracture if high tensile stresses at the underhead radius result from extreme contact
loading conditions in combination with rail material and head wear. It was predicted
that the presence of high underhead radius stresses could lead to catastrophic rail failure
as a result of the rapid fracture of pre-existing RCF cracks extending to the underhead
radius (UHR) region.
The work also examined the rapid fracture behaviour of multiple RCF cracks. A
significant reduction in stress intensity factors, especially for KI and KII, and
consequently in Keq was predicted for multiple cracks compared to a single crack. The
thesis examined the influence of wheel-rail interaction and rail head wear on the
probability of rail failure from the existing gauge corner cracking. It also demonstrated
the differences in rapid fracture behaviour between a single and multiple cracks. The
results were presented for the bending stresses, as they are likely to be different to the
highly localised contact stress driven phases of the āWhole Life Rail Modelā. Therefore,
further investigation of multiple gauge corner cracks in ratchetting and contact stresses
phases would be valuable.
Chapter 9. Conclusions and future work
191
Chapter 9
Conclusions and future work
9.1 Conclusion
The effect of the lateral bending of a whole rail profile and the localised vertical and
lateral bending of the head-on-web has been investigated in order to understand the
phenomenon of fatigue cracking and rapid fracture. Finite element modelling and
extended finite element method (X-FEM) modelling approaches were used. In
particular, the depth at which tensile longitudinal stresses occurred was evaluated, in
relation to the development of the longitudinal stress peak associated with local bending
directly under the wheel position, but this does not need to be the depth of the mode
transition. The magnitude of tensile bending stress was much higher than that associated
with rail uplift ahead and behind the wheel contact (as was identified in the post-
Hatfield studies). The current study only features a single load. To obtain the full effect
of (relative) uplift, a bogie loading should be considered and a vehicle dynamics study
should be required to determine the L/V ratio, contact patch offset and direction of
lateral loading. The results have been described and discussed in detail in each chapter
(4 to 8). This chapter highlights the main conclusions that have been drawn in each
chapter from this research.
The finite element commercial package ABAQUS 6.11-2 was used to examine the
stress state at the underhead radius of the rail. To validate the predicted results, a
comparison with measured field data was conducted with the focus on the longitudinal
stresses at the underhead radius of the rail. The tension spike at the underhead radius
was found to be highly dependent on several factors: the contact patch offset (CPO), the
ratio of lateral (L) to vertical (V) loads, the direction of lateral traction, the vertical
Chapter 9. Conclusions and future work
192
foundation stiffness and the seasonal temperature. The wheel-rail contact location is a
critical factor that influences the stress state at the rail underhead radius, particularly
when offset laterally towards the gauge side. The lateral shear traction generated by
vehicle curving and hunting operations greatly influences the stress state at the rail
underhead radius. Foundation stiffness also has an effect: the stiffer the track the higher
the underhead radius stresses. The magnitude of longitudinal tensile stress increases at
the underhead radius as the vertical foundation stiffness increases. The results also
suggest that the magnitude of longitudinal tensile stress at the underhead radius
increases in the colder months and decreases in the warmer months. This is the result of
rail temperature variations around the nominal stress-free or neutral temperature. This
indicates that stiffer tracks operated in cold weather conditions experience increased
underhead radius stresses.
The tension spike at the underhead radius of a rail increases when the contact patch
moves away from the rail centreline and / or the L/V ratio increases. The direction of
lateral shear traction can be either outward or inward. Inward shear traction was found
to be more damaging. This is easy to understand by considering the effect of L and V on
the rotation and lateral deformation of the rail head-on-web. A contact patch offset
towards the gauge corner rotates the rail head in the same direction due to eccentric
load. The lateral load, L, causes further rotation and lateral deformation of the rail head-
on-web. When the lateral load is towards the gauge corner (inwards), both rotations
(due to L and V) are summative and provide a much larger rotation and lateral
deformation than when the lateral load is outwards. In this case, the rotation due to L
partly cancels out the rotation due to V, producing significantly lower deformation and
longitudinal stresses at the field side underhead radius position. The depth below the rail
running surface at which stress becomes tensile decreases as the L/V ratio increases. For
example, when the L/V=0.4, the depth may be of the order of a few millimeters.
Several new finite element (FE) models were developed to study the effect of worn
profiles, using the commercial finite elements code ABAQUS (6.11-2). The stress state
at the rail underhead radius is influenced by the head wear levels in conjunction with the
wheel - rail contact conditions and L/V ratio. The rail underhead radius with a heavily
worn rail presents the most critical case with very high tensile stress close to the contact
Chapter 9. Conclusions and future work
193
running surface. These high local bending stresses can potentially initiate a fatigue
crack or cause existing rolling contact fatigue cracks to turn downwards. It was found
that the residual and thermally induced stresses interact with these high local bending
stresses and alter the cracking behaviour in a complex way. This can result in a
transition to Mode I crack growth and the formation of a transverse defect for rolling
contact fatigue initiated cracks that extend to this region, as has been observed in
practice [1-4]. RCF cracks generally started to turn down at about 4-5 mm below the
rail head surface for the AS1085.1 [80] head hardened and low alloy heat treated rail
grades operated under high axle load conditions [1, 2, 4] and this depth is dependent on
the rail manufacturing and operational conditions. These tensile stresses may also result
in fatigue crack initiation at the underhead region, as was previously investigated in
connection with the reverse detail fracture in the United States [6, 7, 9].
The results also suggest that stiffer tracks are more prone to high tensile stresses at the
rail underhead radius in the presence of heavily worn rail. The predicted conditions
were in line with field observations of reverse detail fracture defects in the North
American railroad systems [3-10]. Reverse detail fracture defects have been observed
on heavily worn curved rails on stiff tracks subjected to high axle loads. With no
support to the rail from the adjacent sleepers (a vertical foundation stiffness of zero), the
maximum longitudinal tensile stress decreases at the underhead radius but increases at
the base fillet. This condition simulates the effect of ballast pumping, which can remove
support from one sleeper.
A fatigue damage prediction analysis for the rail underhead radius was undertaken to
investigate the fatigue damage behaviour as a function of wheel ā rail contact
conditions, head wear and rail material grades under heavy haul operations. Fatigue
damage was predicted using the Dang Van criterion. Lateral offsets in the wheel contact
patch relative to the rail centreline, in combination with the increased lateral forces
associated with vehicle curving, increase the probability of fatigue damage initiation at
the underhead radius. This is due to an increase in the tensile stresses at the underhead
radius, which is induced by the local bending behaviour on the rail web. In the presence
of heavily worn rail profiles, the fatigue life of rail in sharp curves is less than in tangent
Chapter 9. Conclusions and future work
194
track. In general, the more wear that occurs on the surface of the rail, the more fatigue
damage is predicted to initiate at the rail underhead radius.
The analyses revealed that, under severe loading conditions, fatigue damage is predicted
to develop at the rail underhead radius with increased head wear and in a rail grade with
lower fatigue limit. It was reported in the North American railroad systems that a
reverse detail fracture is initiated in poorly lubricated, heavily worn curved rail on stiff
track carrying traffic under high axle load conditions, with flow lips as initiators [6-10],
and additionally there is a possible contribution of higher residual stresses at a rail weld.
Failures that initiate at the underhead radius of the aluminothermic welds have been
found in Australia on lines subjected to heavy axle load railway operations. However,
for high strength rail material with higher fatigue limits, it was rarely predicted that
fatigue damage would occur at the underhead radius under the range of loading
conditions considered with an increase in the extent of the head wear. Examples of the
high strength rail materials used included the low alloy or hypereutectoid heat-treated
grades used in Australian heavy haul operations. The results of the fatigue analysis
model showed sufficiently good correlation with field observations.
The prediction of rapid fracture behaviour of pre-existing RCF cracks that have already
turned down (to Mode I behaviour), associated with the lateral bending of the whole rail
profile and localised vertical and lateral bending of the rail head-on-web, was described
in chapters 7 and 8. In particular, the rapid fracture behaviour of long RCF cracks in
phase 3 of the āWhole Life Rail Modelā was considered. The extended finite element
method (X-FEM) was used to investigate the stress intensity factor (KI, KII and KIII) near
the crack front that results from contact and bending stresses. Several new crack models
were developed by inserting X-FEM elements in the existing finite element (FE) meshes
for the prediction of RCF cracks unstable growth behaviour under mixed-mode loading
at the gauge corner of the rail. Both single and multiple cracks were modelled at different
crack lengths and orientations, over a range of head wear conditions.
The results of rapid fracture analysis using mixed-mode loading suggest that when a
wheel passing on the rail head without eccentricity is above the crack mouth, the head is
in compression due to global bending. In the plane tensile opening mode, KI is negative
and the Keq values are very small, far below the fracture region defined by Keq KIC.
Chapter 9. Conclusions and future work
195
The underhead radius stresses are higher in heavily worn rails. An increase in the
contact patch offset increases these stresses and stress intensity factors even more. The
effect of KI max influenced by the contact patch offset is a case in which high tensile
bending stresses are produced at the underhead radius due to the localised bending of
the rail head-on-web and lateral bending of the whole rail profile. As a result of changes
in the stress state, the KI increases and so does the Keq. The results revealed that the
highest KI max occurs when the contact loading is directly above the crack location and is
exacerbated by increasing head wear.
The L/V ratio also plays a significant role in the tensile bending stresses at the
underhead radius. The highest tensile bending stress at the underhead radius is caused
by the superposition of the eccentric loading with inward lateral shear traction in
combination with an increase in head wear when the contact loading is applied directly
above the crack location. This leads to the higher in-plane opening stress intensity factor
KI at the 0 crack front position, corresponding to the underhead radius near the lower
gauge corner point. Similarly the Keqmax is highest at the same location due to the large
contribution of the positive in-plane tensile opening KI. The results revealed that the in-
plane tensile opening stress intensity factor KI is the dominant driver of a fracture.
Nevertheless, the L/V ratio also has a significant effect on in-plane sliding KII mode and
consequently Keq when the contact position is away from the crack. The values of the
stress intensity factors in the out-of-plane tearing KIII mode are much lower than the
opening KI and the sliding KII modes, suggesting that the in-plane tensile opening and
shear forces play an important role in the unstable propagation and extension of pre-
existing RCF cracks to the rail underhead radius. Therefore, KI max and KII max were more
significant in terms of rapid fracture than KIII max.
The results revealed that a contact patch offset with a L/V ratio could drive the rapid
fracture of pre-existing RCF cracks at the gauge corner region, with the crack length
extending to the rail underhead radius, if a heavily worn rail exists in practice for the
combination of rail material grades considered. In practice, most of the cracks occur on
the gauge corner of a high rail with curves of decreasing radii, around 850 m ā 350 m,
as was reported by Kerr et al [193] and these are more prone to rail fracture. The curved
rails are generally heavily worn, with contact patch eccentricity due to steering effects
Chapter 9. Conclusions and future work
196
and cant deficiency compared to tangent tracks. In addition, the curves are prone to high
lateral shear traction, which may increase the risk of rail failure due to the rapid fracture
of pre-existing RCF cracks.
An examination was also made of the rapid fracture behaviour of multiple RCF cracks.
A significant reduction in stress intensity factors, in particular, for KI and KII, and
consequently in Keq is predicted for multiple cracks compared to a single crack. It has
been commonly observed that only a small number of widely spaced cracks in the series
can become longer compared to others. In this way, there is a possibility that the
shielding effect may be reduced and a single crack may become critical, resulting in the
rapid fracture of the rail, as presented in the case of single crack models. The single
crack case is more critical and the analysis and design against failure of a single crack
should be considered accordingly. The results of this work can be used to examine the
influence of wheel-rail interaction behaviour and rail head wear on the probability of
catastrophic rail failure resulting from long turned down pre-existing RCF cracks. The
thesis has also demonstrated the differences in the rapid fracture behaviour between
single and multiple cracks. It is suggested that KI and KII should be primarily reduced to
prevent rapid fracture.
The analyses results reported in this thesis have applications in the improvement of the
modelling of RCF crack propagation and fatigue damage prediction at the rail
underhead radius, leading to improved asset management and risk assessment. The
approach taken in this thesis forms a basis for more extensive studies into the growth
behaviour of RCF cracks, as outlined above, in order to be better able to provide firm
guidelines to rail management, particularly in relation to RCF damage. The following
measures are proposed in this regard:
Catastrophic rail failure due to the rapid fracture behaviour of pre-existing long
turned down (to Mode I behaviour) RCF cracks can be prevented by following safe
head wear limits through appropriate maintenance strategies. It is recommended that
the approach studied in relation to RCF cracks be used when determining safe head
wear levels.
Chapter 9. Conclusions and future work
197
Rail life can be improved by reducing the contact patch lateral displacement. Some
of the approaches include but are not limited to re-profiling the rail by appropriate
maintenance and grinding strategies.
Rail performance can be improved by controlling the L/V ratio through the
application of lubrication or other friction management strategies (friction modifier)
at the wheel - rail interface, particularly in short radius curves.
In addition to lubrication strategies, the bogie suspension design should be
optimised to control the lateral loading, especially the inward lateral shear traction,
which is the more damaging phenomenon. Nevertheless, more rigorous studies are
required to work out a strategy as to how this problem can be solved.
As far as possible, sharp curves should be avoided. There are several examples of
curve radius reductions in connection to speed increases of existing railway lines.
This could also be done when new routes are planned.
Stiffer tracks should be avoided, as the tensile bending stresses increase at the
underhead radius and leads to a higher propensity for rapid fracture. Therefore, it is
beneficial to keep the foundation stiffness (sleeper type, fasteners, rail sleeper pad
and ballast) within the design range through regular maintenance. A revision of
asset management procedures can also be considered.
Fatigue crack initiation at the underhead radius can be prevented by using rail steels
with high fatigue limit, such as low alloy or hypereutectoid heat treated grades. High
strength steels tend to be more sensitive to surface roughness and geometric defects
and the risk of a rapid fracture of these high strength rail materials should be
balanced with their fracture toughness āKICā
9.2 Suggestions for further work
The inferences drawn from analyses results are true for the limited combination of
operating conditions and rail material grades considered. Further detailed analyses and
field investigations are required to understand how the combinations of operating
conditions and rail material grades affect rolling contact fatigue cracks and head wear.
The recommendations for future work are summarized in this section.
Chapter 9. Conclusions and future work
198
The fatigue life model results indicate under which combination of loading conditions,
rail head wear and rail material grades fatigue damage is predicted to be initiated.
However, in practice, fatigue failure is generally associated with the presence of pre-
existing defects or a stress concentrator in the form of a sharp radius, which may include
flow lips (as was evident in the reverse detail fracture failure mode) or rolling contact
fatigue cracking. The simulation results for fatigue damage prediction did not consider
the effects of pre-existing defects or other conditions, which may increase localised
stresses in the vicinity of the underhead radius. However, the results do indicate the
potential formation of defects such as reverse detail fractures, which have previously
been found to develop at the underhead radius (lower gauge corner) on heavily worn
rail. Further work is required to examine these behaviours in the presence of pre-
existing defects.
The residual stresses were incorporated using results obtained by the neutron diffraction
method. These values may not be representative of current rail manufacturing
procedures, and hence the treatment of residual stress aspects needs to consider the fact
that those in the rail grades examined may differ from those used for the previous
neutron diffraction measurements. The approach presented can also be extended to
include the original residual stress distribution for both roller-straightened rail, and that
straightened by other methods, such as stretch straightening. The influence of pre-
existing residual stresses resulting from the rail manufacturing process could be
examined by undertaking a sensitivity analysis to residual stresses. In addition, the
repeated rolling contact between wheel and rail is another important parameter that can
lead to a complex distribution of the residual stresses in the rail and is of interest for
future study.
The vehicle dynamic effects associated with variations in track geometry will also
increase the magnitude of the vertical and lateral loads. The dynamic effects will change
the contact patch offset and L/V ratio in addition to the increased load. Increased train
speeds may also produce an increase in dynamic loads as well as an increase in the
effect of lateral forces due to lateral irregularity on either the gauge or the field side, and
subsequently an increase in the L/V ratio. The current analysis did not consider vehicle
dynamics. This model can be updated by using the contact load conditions (contact
Chapter 9. Conclusions and future work
199
patch size, shape and position, L/V ratio etc.) based on vehicle dynamic studies to
obtain a detailed prediction of the stress state under dynamic loading conditions and
would be of interest for future study.
The fatigue crack growth rate and the direction of the propagation of pre-existing RCF
cracks using the contact patch offset, lateral traction and head wear conditions are still
unclear. Further study is required to model the quantitative assessment of fatigue crack
growth using more realistic data, both in field and laboratory conditions. Moreover,
water entrapment and crack closure are factors that are not considered in the extended
finite element method (X-FEM) fracture modelling. Further development of the current
model could include crack closure and liquid entrapment, in addition to localized
bending effects. The behaviour of multiple RCF cracks also requires examination under
more realistic service conditions. The results of rapid fracture analysis of multiple RCF
cracks were presented for phase 3 of the āWhole Life Rail Modelā, mainly driven by
bending stresses that are likely to be different to highly localised contact stress driven
phases. Therefore, further investigation of multiple RCF cracks in ratchetting and
contact stress phases would be valuable.
Rapid fracture analysis performed for pre-existing RCF cracks incorporated fracture
toughness KIC values based on previous studies (as reported in the respective
references), and was used solely to illustrate the effect of differences due to rail grade.
Newly introduced rail steels, such as bainitic grades, also have the potential to replace
the widely-used pearlitic steels. A more rigorous analysis would require testing of the
rail material grades under identical test conditions, to ensure that any differences in KIC
were not attributable to the test methods used. Therefore, a thorough review of the rail
steel grades should be considered and incorporated into a future study.
The prediction of safe head wear limits was based on the limited head wear profiles
considered. Further study requires multiple sets of analyses with a range of head wear,
crack size, orientations, separation and loading conditions, in order to see the sensitivity
of these limits in detail. The approach presented in this thesis, if extended to include the
residual stress aspects in the presence of pre-existing defects or the other conditions
outlined above, provides a basis for an assessment of allowable rail wear limits as a
Chapter 9. Conclusions and future work
200
function of loading conditions and rail material grades and may be worth investigating
in a future study.
References
201
References
1. Mutton PJ, Tan M, Bartle P, Kapoor A. The effect of severe head wear on rolling
contact fatigue in heavy haul operations. 8th International Conference on Contact
Mechanics and Wear of Rail/Wheel Systems Comitato (CM2009); September 15-
Appendix C. FORTRAN-code based on the Dang Van fatigue criterion
SUBROUTINE UVARM(UVAR,DIRECT,T,TIME,DTIME,CMNAME,ORNAME, 1 NUVARM,NOEL,NPT,LAYER,KSPT,KSTEP,KINC,NDI,NSHR,COORD, 2 JMAC,JMATYP,MATLAYO,LACCFLA) C INCLUDE 'ABA_PARAM.INC' C CHARACTER*80 CMNAME,ORNAME CHARACTER*3 FLGRAY(15) DIMENSION UVAR(NUVARM),DIRECT(3,3),T(3,3),TIME(2) DIMENSION ARRAY(15),JARRAY(15),JMAC(*),JMATYP(*),COORD(*) C CC Error counter: JERROR = 0 CC C CALL STRESS STATE FROM EACH POINT IN THE DEFINED SETS C============================================================================ C Stress tensor: FOR EACH TIME INCREMENT AT THE MATERIAL POINT CALL GETVRM('S',ARRAY,JARRAY,FLGRAY,JRCD,JMAC,JMATYP, 1 MATLAYO,LACCFLA) C JERROR = JERROR + JRCD SIG11=ARRAY(1) SIG22=ARRAY(2) SIG33=ARRAY(3) SIG12=ARRAY(4) SIG23=ARRAY(5) SIG31=ARRAY(6) C Press stress:FOR EACH TIME INCREMENT AT THE MATERIAL POINT CALL GETVRM('SINV',ARRAY,JARRAY,FLGRAY,JRCD,JMAC,JMATYP, 1 MATLAYO,LACCFLA) Mise=ARRAY(1) Shh=ARRAY(3) INV3=ARRAY(4) C Principal stress:FOR EACH TIME INCREMENT AT THE MATERIAL POINT CALL GETVRM('SP',ARRAY,JARRAY,FLGRAY,JRCD,JMAC,JMATYP, 1 MATLAYO,LACCFLA) JERROR = JERROR + JRCD SP1=ARRAY(1) SP2=ARRAY(2) SP3=ARRAY(3) C Note (SP1<SP2<SP3) Tmax=0.5*(SP3-SP1) Tmin=0.5*(SP2-SP1) Tmid=0.5*(Tmax+Tmin) C=============================================================================
Appendix C. FORTRAN-code
262
C DEFINING THE DANG VAN PARAMETERS: C Enter the fatigue limit from the tension-compression test (in MPa): Sfl=353.0 C Enter the fatigue limit from the torsion test (in MPa): Tfl=205.0 aDV=3*((Tfl/Sfl)-0.5) bDV=Tfl C START TO CHECK SHEAR PLANE Do I=1,360 C LL START TO CHECK EVERY 1 DEGREE IN THE Nx DIRECTION Do J=1,360 DO K=1,360 C AT EACH 1 DEGREE OF Nx, TO CHECK ALL DEGREES OF Ny DIRECTION at=I-1.0 bt=J-1.0 ct=k-1.0 aL=COSD(at) aM=COSD(bt) aN=COSD(ct) Tal=aL**2+aM**2+aN**2 If (Tal.eq.1.0) then C TOTAL STRESS IN āNā DIRECTION Snn=SIG11*aL**2+SIG22*aM**2+SIG33*aN**2+2*(SIG12*aL*aM+ +SIG23*aM*aN+SIG31*aL*aN) C Sxx=SIG11*aL+SIG12*aM+SIG31*aN Syy=SIG12*aL+SIG22*aM+SIG23*aN Szz=SIG31*aL+SIG23*aM+SIG33*aN Tsp=(abs(Sxx**2+Syy**2+Szz**2-Snn**2))**0.5 C DANG VAN SHEAR STRESS Tt=ABS(Tsp-Tmid) BB=Tt+abs(aDV*Shh) If (BB.GT.Tfl) THEN Du=5*(Tfl-BB)/(Tfl-Sfl)-6 Dl=10**Du aNf=1/Dl UVAR(1)=BB UVAR(2)=ACOSD(aL) UVAR(3)=ACOSD(aM) UVAR(4)=ACOSD(aN) UVAR(5)=Tsp UVAR(6)=Tt UVAR(7)=Dl UVAR(8)=aNf
Appendix C. FORTRAN-code
263
Else UVAR(9)=BB ENDIF EndIF Enddo ENDDO ENDDO write(*,*)'====================================' write(*,*)'Time INC=',DTIME,'Total time=',time(2) write(*,*)'ELE No:',NOEL write(*,*)'Tsp=',Tsp write(*,*)'Tmax=', Tmax write(*,*)'Dang Van Shear_BB=',BB write(*,*)'Plane n to X axis=',UVAR(2),'Deg' write(*,*)'Plane n to Y axis=',UVAR(3),'Deg' write(*,*)'Plane n to Z axis=',UVAR(4),'Deg' C If error, write comment to .DAT file: C IF(JERROR.NE.0)THEN C WRITE(*,*) 'REQUEST ERROR IN UVARM FOR ELEMENT NUMBER ', C 1 NOEL,'INTEGRATION POINT NUMBER ',NPT C ENDIF RETURN END