Wheel-Rail Contact: GETTING THE RIGHT PROFILE

Transcription

1 Wheel-Rail Contact: GETTING THE RIGHT PROFILE Simon Iwnicki, Julian Stow and Adam Bevan Rail Technology Unit Manchester Metropolitan University The Contact The contact patch between a wheel and a rail is typically the size of a 5p piece and its shape depends on the geometry of both bodies. 2a 2b The theory developed by Hertz (when he was a 24 year old research assistant!) can be used as a good approximation and this gives an elliptical shape and distribution of the normal force. 1

2 Alternative contact models In practice the geometry at the contact patch is often more complicated and the contact patch is far from elliptical. Possible alternative models are: Hertz theory in strips Finite elements Can give good results but very slow FE model of wheel rail contact Courtesy Tom Kay Corus Rail technologies 2

3 The Forces Vertical / Normal: (the vehicle load) P0, P1 and P2 forces Lateral / Tangential: (traction, braking, curving) Gravitational stiffness forces Creep forces Creep forces A tangential force related to creepage or microslip in the contact patch Creep force Linear Nonlinear Saturated Creepage Pure rolling Full slip 3

4 Alternative creep force models Carter linear relationship between creepage and creep force Johnson and Vermeulen modification due to saturation Kalker theoretical full solution Shen Hedrick Elkins refined and simplified based on measurements Pollach useful for high slip Falling creep force essential for stick-slip and squeal Wheel and rail profiles wear and their geometry changes from the original design. The Profiles Miniprof measured rail profile Profiles can be measured and combined to produce the effective conicity or a table of radii and contact angles. 4

5 Measuring profiles Simulation of contact 5

6 Simulation of forces Contact - animation 6

7 Current profiles Tread cone angle + Flange angle + Radius egp1 Worn profile eg P8, S1002 (P10) The UK P8 - a worn profile 7

8 Tools for designing new profiles Rules of thumb Flange angle Cone angle Geometry analysis contact location Rolling radius difference plot curving, stability Shakedown curve Rolling Contact Fatigue Wear prediction profile life Geometry analysis

9 Rolling radius difference Low Conicity Wheelset rolls offset From centre of track Sharp Conicity Increase RCF Assessment Two current methods: - Shakedown curve (more later) - Weighted Tγ 9

10 Wear models T gamma model 1 µ Wn = Tyγ y + T A 0.6 [ γ ] x x where: Wn = the wear number A = the contact patch area µ = the coefficient of friction between the wheel and rail Tx,Ty = longitudinal and lateral creep forces γ x γ y = longitudinal and lateral creepages Wear models Archard wear model: V = k 1 F n s/h (m 3 ) where k 1 is a non-dimensional wear coefficient, H is the hardness of the softer material, F z is normal contact force, s is the sliding distance in the contact patch. 10

11 Local wear rate in the contact patch (Archard model) Wear /m 2 x sliding Slip zone wear 1.5 δz Adhesion adhesion zone no wear Rolling direction 0 -b 0 y b a x 0 -a Simulated and measured wheel profile and wear distribution example 1 -after km 10 Simulated Distance 54210km (mm) Simulated -30 Measured New Wear Distribution(mm) Simulated Measured (mm) 11

12 Simulated and measured wheel profile and wear distribution example 2 - after km 10 Simulated Distance km (mm) Simulated -30 Measured New Wear Distribution(mm) Simulated Measured (mm) An Example: The WRISA2 anti-rcf wheel profile (development funded by RSSB) 12

13 Contact position simulation for c2c East Ham Depot - Contact Stress vs Position class 357 class 312 S i Contact Stress / MN/m^ Distance /mm Contact Conditions P8 Profile 13

14 Contact Conditions P1 Profile No contact in these positions New P1 Wheel on New BS113a Rail Heavy Worn P1 Wheel on Worn BS113a Rail P1 wheel on c2c 14

15 The WRISA2 anti-rcf profile Anti-RCF Relief WRISA2 profile designed by NRC to: Include an Anti-RCF relief in flange root area of the profile Reduce the contact stress in mid-gauge region of rail Reduce the rolling radius difference when curving to reduce longitudinal creepages and tangential forces Include a smoother tread run-off to eliminate geometric stress raisers caused by transition. RCF Sensitive Region WRISA2 Rolling Radius Difference New P1 (Red), New P8 (Green), Light Worn P8 (Blue), WRISA2 (Purple) 15

16 WRISA2 - Shakedown Exceedence How does Nature solve these problems? (Which is the best beak?) 16

17 Genetic Algorithm Basic Method - (simulates natural evolution) Profile digitised and converted to a binary gene 2 parent profiles selected and their genes mated to produce child genes Child genes converted back to profiles Simulation used to evaluate each child profile Best profiles selected as parents and process repeated More details in Persson and Iwnicki paper presented at 18 th IAVSD conference 2003 The genes x: mm y:5.223mm x: mm y:6.201 mm X:

18 Mating Bits are copied with random switches between the parents = Reconstructing the new profile x: mm y:5.223mm x: mm y:6.201 mm X:

19 Survival of the fittest Each profile is assessed by running a vehicle simulation over a typical track case. The best profiles become the parents for the next generation. Mutations can be added to avoid local minima. Some interim results 19

20 Assessing the new profiles The Penalty Index - the sum of sub factors: Ride comfort penalty factor Lateral track shifting force penalty factor Maximum derailment quotient penalty factor Wear penalty factor Maximum contact stress penalty factor (Other factors are possible and each can be weighted according to importance) Results - stiff bogie 20

21 Results - soft bogie Conclusions The design of wheel profiles for the conflicting requirements of stability and derailment resistance as well as low wear and resistance to rolling contact fatigue is a significant engineering challenge Rules of thumb related to tread cone angle and flange angle have limited usefulness Various tools are available to predict contact conditions and possible problems such as RCF and wear A novel method using a genetic algorithm together with a penalty index and a dynamic simulation can help to optimise profiles for a particular vehicle or track case. 21