Académique Documents
Professionnel Documents
Culture Documents
TOWARDS FAST AND ACCURATE MODELLING OF
WHEEL‐RAIL CONTACT
Matin S. Sichani, Roger Enblom, Mats Berg
Theory Trade‐off between accuracy and time efficiency
Results
• Elliptic models:
Conclusions
o Hertz,
Hertz, equivalent elastic
equivalent elastic
o Non‐physical patch estimation and inaccurate stress distr. in several
contact cases
• Non‐elliptic contact models:
o Advanced models: using FEM or BEM
o Fast models: simplified contact conditions
Fast models simplified contact conditions
• Analytical estimation of the contact patch
• Based on virtual interpenetration (VI)
I t d ti
Introduction
Theory • Tangential part based on FASTSIM (Kalker 1982)
Results
• Well‐known models:
Conclusions
o Kik
Kik‐Piotrowski
Piotrowski (1996)
(1996)
o Linder(1997)
o Ayasse‐Chollet (2005) (STRIPES)
NSJ 2012–Tammsvik 3
Virtual interpenetration
• The exact contact equation:
Penetration
I t d ti
Introduction Normal
Normal
Separation
deformation
Theory
Results
• Assume a rigid interpenetration:
Assume a rigid interpenetration:
Conclusions
• Penetrate by a fraction of the prescribed pen.:
Penetrate by a fraction of the prescribed pen :
• Solve the new equation to find the patch boundaries:
Solve the new equation to find the patch boundaries:
NSJ 2012–Tammsvik 4
Linder method
• The scaling factor is set to a constant:
• In a Hertzian case, the semi‐axes are:
Curvatures at the point
of contact
Introduction Hertz
Linder
Theory • The Hertz solution is:
Results
Conclusions
• The semi‐axes ratio:
Geom. coeff. based
on curvatures
• Same results, only if:
NSJ 2012–Tammsvik 5
Kik-Piotrowski method
• Same scaling factor but a correction is applied
• Axes ratio is set equal to Hertz’s:
Corrected semi‐axes
Introduction Hertz
Kik-Piot
Theory • The semi‐axes are then:
Results
Conclusions
• Area may still be different from Hertz’s!
y
NSJ 2012–Tammsvik 6
STRIPES (A-correction)
• Scaling factor is a constant based on the geometry:
Corrected curvature at yi
Corrected curvature at y
Introduction • If only the long. curvature is corrected:
If l th l t i t d Hertz
STRIPES
Theory
Results
• Semi‐axes are:
Conclusions
o exactly the same as Hertz’s !
NSJ 2012–Tammsvik 7
STRIPES (A&B-correction)
• Scaling factor is based on the geometry:
• Both curvatures are corrected:
Introduction Hertz
STRIPES
Theory
Results
• Semi‐axes are:
Semi axes are:
Conclusions
o Same length but different width from Hertz’s!
• Approaches Hertz’s when → 0 or 1
NSJ 2012–Tammsvik 8
Tangential part
• Tangential part is treated using FASTSIM
• FASTSIM is originally used for elliptical patches
Introduction • Equivalent ellipse:
E i l t lli
Theory o Kik and Piotrowski (1996)
o Defining an equivalent ellipse for each zone, using elliptical flexibility
Defining an equivalent ellipse for each zone using elliptical flexibility
Results
parameters
Conclusions
• Local ellipses for each strip:
o Linder (1997) , Ayasse and Chollet (2005)
o Strip discretization of the patch
S i di i i f h h
o Assigning an ellipse into each strip and using its elliptical parameters
NSJ 2012–Tammsvik 9
Local ellipse assignment
• Linder method:
o All local ellipses have the same
Introduction lateral semi axis
lateral semi‐axis.
Theory
Results
Conclusions
• STRIPES method:
o Local ellipses are based on the
local curvature values at the
centre of the strip
Towards gauge‐corner
NSJ 2012–Tammsvik 10
Case study
• Models are implemented using MATLAB
o As documented in the literature
o STRIPES model: A
STRIPES model: A‐correction
correction approach
approach
• A single wheel on rail example is solved:
Introduction
o Right wheel‐rail pair
Theory o S1002/UIC60 (1:40)
o Zero lateral displacement (central wheelset position)
Zero lateral displacement (central wheelset position)
Results
o Normal load: 78.5 kN
Conclusions o Spin= 0.052 rad/m (pure spin)
p / (p p )
• CONTACT software (BEM) results are taken as reference
o Since, it is bound to half‐space assumption, this case study is confined
to tread contact
NSJ 2012–Tammsvik 11
Contact patch
Introduction
Theory
Results
Conclusions
Towards gauge‐corner
NSJ 2012–Tammsvik 12
Contact pressure distribution
Introduction
Theory
Results
Conclusions
NSJ 2012–Tammsvik 13
Creep forces
Introduction
Theory
Results
Conclusions
CONTACT: same spin throughout the patch
p g p STRIPES: spin calculated at
STRIPES: spin calculated at
the center of each strip
NSJ 2012–Tammsvik 14
Improving pressure distribution
• In STRIPES, negative B values
are not allowed
are not allowed.
o The negative B is replaced by
a minimal positive value
Introduction o Smoothing filter is then applied
to achieve smooth patch boundaries
Theory
Results
Conclusions
• Smoothing also affects the pressure values and smears out the peaks
• To avoid smoothing, a new correction of negative B values is suggested
o Instead of cutting negative B
Instead of cutting negative B values out, shift them upward
values out, shift them upward
o Avoid sharp changes in B
• Since A‐correction is sensitive to low B values, A&B‐correction is also of
i
interest.
NSJ 2012–Tammsvik 15
Wheelset central position
Introduction
Theory
Results
Conclusions
CONTACT
STRIPES
STRIPES (A&B-correction)
NSJ 2012–Tammsvik 16
Off-set case: Δy = -1 mm
Introduction
Theory
Results
Conclusions
CONTACT
STRIPES
STRIPES (A&B-correction)
NSJ 2012–Tammsvik 17
Off-set case: Δy = 1 mm
Introduction
Theory
Results
Conclusions
CONTACT
STRIPES
STRIPES (A&B-correction)
NSJ 2012–Tammsvik 18
Off-set case: Δy = 2 mm
Introduction
Theory
Results
Conclusions
CONTACT
STRIPES
STRIPES (A&B-correction)
NSJ 2012–Tammsvik 19
Conclusions
• Patch prediction by VI‐based models should be improved in off‐set
cases
Introduction • Contact pressure distr. deviates from reference in wheelset central
Contact pressure distr deviates from reference in wheelset central
position
Theory
Results
• The A&B‐correction strategy and avoidance of smoothing leads to
Conclusions improved pressure distr. prediction
• Non‐planar patch and spin variation considered by STRIPES.
P
Pronounced effects in the gauge‐corner contact
d ff t i th t t
NSJ 2012–Tammsvik 20
Future work
• Comparison of fast non‐elliptic models to FEM results in gauge‐
corner contact
t t
Introduction
Theory
• Improve the contact patch estimation of models based on virtual
interpenetration (in off‐set cases)
p ( )
Results
Conclusions
• Investigate the non‐elliptic model based on semi‐Winkler approach
(Telliskivi 2004)
NSJ 2012–Tammsvik 21
Thanks for your attention!
A
Any questions or comments?
ti t ?
Matin S. Sichani
Email: matinss@kth.se
22
17th Nordic Seminar on Railway Technology - Tammsvik