Below are reproduced the questions posed by the audience and the answers given by the speakers at the CM2003 conference in Gothenburg 10 -- 13 June 2003. Please note that transcription of the handwritten notes may have introduced errors. If you find an error in any of your questions or answers, please e-mail us on info_at_charmec.chalmers.se.
What remedial measures are you proposing to alleviate the severe type of RCF damage observed?
Results show that the severe situation is perhaps the result of a combination of factors: (i) New vehicles, with small radius wheels and unequal weight distribution. (ii) Vehicle speed. (iii) Rail material. The type of analysis set-up perhaps allow to distinguish the influence of the different factors (e.g. fixing and max speed limit depending on wheel radius as well as restriction to max allowable axle loads). Perhaps the main result is that ratchetting rate is higher than wear in low radius curves. A preventive action is surely a periodic grinding able to remove the "ratchetting" (or RCF) affected zones.
Are the extreme magnitudes in the shakedown diagram mainly due to a very small contact patch or very high contact forces?
In general contact forces are lower than in railway systems, but the main feature of contact patches for tramways is that they are very narrow, especially at the head corner, at the flange and at the counter-rail (due to flange-back contact). The important factor leading to very high ratchetting intensities is the full-slip regime in low rolling curves.
1) How did you define failure?
2) How big were the material defects found?
1) Failure was defined as a 10% drop in axial or torsional stiffness.
2) The material defects found had an average length of about 20 micrometers.
How are rotational irregularities of the roller rig brought into account for the lateral force measurement method?
Rotational irregularities are measured on unloaded roller and are then subtracted from the measure in the loaded case.
1) Did you calibrate for varying track gauge? If not, did you analyse it?
2) The vertical force actuators are positioned above the suspension. Up to what frequency do you consider the actuator force to be equal to the contact force?
1) The roller rig does not allow gauge variation, so the wheel-set was calibrated for one single gauge value. I expect vertical and longitudinal forces to be insensitive to gauge variation. For lateral forces further analysis is required.
2) Besides actuator forces, also the accelerations of the semi-bogie and wheel-set are measured. So in principle, we should be able to measure correctly the forces in the 0--30 Hz frequency range required by the standards. Anyway, until now the experiments were limited to a much lower frequency range.
What was the test speed in the dynamic calibration? Do you expect the calibration to vary with speed?
The speed was 100 km/h and was
not changed. In case the
calibration turns out to be speed dependent, it would be easy to define
different calibration matrices for different speeds.
You have obviously large activities in this field at INSA Lyon. How many persons are working with problems regarding wheel--rail interaction?
We are currently 8 persons involved in wheel--rail contact problems in the Tribology team at the LaMCos, INSA.
1) There are similarities in behaviour of interface layers in wheel--rail contact and rolling of steels. Have the authors considered this?
2) Micrographs show regions of white etching layers. Could you comment on this formation?
1)The wheel--rail contact and the rolling of steels are two applications of tribology. Both applications present similarities in terms of high pressure coupled with shearing undergone by the "skins" of materials in contact.
2) The formation of white etching layers under contacts are under investigation in our team. So far it appears that these layers form as a consequence of very high pressure coupled with shearing and without a preponderent effect of temperature.
Can 3rd body be treated as a hard layer which gets pushed into the wheel--rail surfaces or as a "lubricant"?
Depending on the operating conditions, the solid layer can play the role of a solid lubricant.
What is the difference between the user defined variable and other strains?
There is no difference user
defined variable is merely visual.
Experiments with oil lubricated discs showed that the friction with transversely ground discs was nearly double of that with circumferentially grinding in the dried lubrication regime.
In EHL theory, if the contact area is perfectly covered with lubricant, friction depends on the thickness of lubricant. In fact, the thicker the film is, the larger the friction is. However, if the contact area is not perfectly covered with lubricant, which means that the mixed lubrication regime can be adopted, friction depends roughly on the solid contact area or asperity contact area, which means the thicker the averaged thickness of the lubricant is, the smaller the friction is. As the authors suggested in this paper, everybody can easily agree with their results that the averaged thickness of lubricant on transversely ground discs is larger than that on circumferentially ground discs. According to the above discussions, your experimental results can be governed by EHL theory. In fact, the contact area of your experiments should have been perfectly covered with oil lubricant, which means mixed lubrication theory can't be adopted in your case.
I'm curious about the software for the multilevel computations:
1) What kind of software do you use?
2) Do you use any standard finite element applications?
1) I referenced the book by Venner & Lubrech, Multilevel Method in Lubrication, Elsevier, 2000, and made some modifications to their program.
2) No. I divided the domain of calculation into rectangular uniform grids in my in-house code.
1) Transverse asperity orientations appear better for adhesion -- are there practical ways of introducing such asperities into the rail or wheel?
2) If so, will they be bad for noise?
1) The analysis solutions that longitudinally oriented roughness is better for adhesion have not been applied to railways yet. However, in practice, the wheels are ground along the circumference direction.
2) I did not investigate the effect of roughness orientation on noise, so I don't know.
How do you take into account frictional forces in the contact area? Are you able to detect the part of the contact patch which is in adhesion and the part where slip occurs?
The finite element method calculates in each time increment (some nano-seconds) the equilibrium and the acting contact forces combined with the Coulomb friction model. The mesh was much too course to investigate stick--slip behaviour in the contact patch, but this is possible using the explicit FE method and a finer mesh.
Your results seem to overestimate LBF results (Grubisic et al). What parameter can you tune for matching these results?
1) I forgot to mention that also the weight of the wheel (1 ton) was added to the 70 kN axle load.
2) In the calculations the elastic part attached to the rigid wheel was too thin, so our results with a fully elastically modelled wheel produces 10--15% lower reaction forces.
Experiments in Cambridge passing small billots of plasticine between rollers have shown that the billot always emerges from the nift adhering to the driving surface. Have you observed this phenomenon in your experiments?
Effectively in the case of tests performed on the simulator the roller is driving. But I also believe that mechanical and physico-chemical conditions should be taken into account. We observe that 3rd body is very sensitive to the environmental and operating conditions. The rheology of the 3rd body changes as a function of these conditions (rheology is characterised by cohesion, ductility and adhesion of the first bodies). As a result, velocity accommodation can take place either in the thickness of the third body or in the skin of the bodies in contact or both.
1) I believe that the structure of the "white phase" you refer to is martensite, what is your opinion?
2) The so-called 3rd body, which structures does it consist of?
1) It is assumed that it is a
martensitic structure.
2) The 3rd body is different from the white phase. It is a mixed layer stemming from the wheels and the rail materials. Its cristallographic structure is undetermined. Some hardness measurements have been performed, but it is still difficult because of the cohesion and discontinuity of this layer.
Is the white phase the same as the white-etching layer? What about the hardness of this layer?
Effectively, the white phase is the "white etching layer". It is linked by epitaxy to the first body (wheel or rail) and so is different from the 3rd body. The hardness of the etching layer is commonly three times higher than the hardness of a rail surface zone, which does not present the white etching layer.
Have you examined the flows of 3rd body material in curves compared to tangent track?
The replica, to highlight 3rd body flows you have seen here, are issued from rails on straight lines. Studies are performed to examine rails in curves. But 3rd body flows can be longitudinal and/or transversal. It depends on the contact conditions. The question could be: "are particles trapped in the contact"!
1) Do sand particles actually stick into the wheel surface and could cause a periodical wear/groove on the rail?
2) Apparently, sand causes great wear and roughness of the rail. In the light of this conference, it seems logical to draw conclusions. Would you consider it wise to look for other adhesion improvers?
1) No, we did nont find any sand particles remaining embedded in the disk after testing. We use the term stick to mean the opposite to slip, rather than "embed".
2) Yes, definitely. Sand will damage the track. But perhaps there is some possible optimization to find a practice that minimizes surface damage.
Is there a scaling effect between the twin disc tests results and full scale testing regarding the size of the particles entering the contact and the relative sliding velocity?
The crushed sand particles are very small compared with the size of the line disc contact (which is similar in size of the wheel--rail contact. The large scale in full scale testing will mean that larger particles can be entrained.
1) Could you further explain the reason for the flange noise reduction due to the friction modifier including the frequency domain 5000--10000 Hz?
2) Could you further explain the experimental conditions of ground-borne vibration and the possible reason behind reducing it due to friction modifier?
1) We believe flange noise reduction is due to reduced lateral and flanging forces. This in turn is a result of controllably reduced friction (0.35) on the top of the rail. Lateral force is mainly influenced by friction on the top of low rail, flanging force by the top of both high and low rail. We are not sure why flanging noise appears to be particularly associated with the 5000--10000 Hz frequency domain.
2) The paper has details on ground-borne vibration. Vertical vibrations were measured 10 m from the track in a small building. At this point we are not sure of the mechanisms of vibration reduction but speculate it is due to better curve performance leading to "smoother " train handling.
Do you have any indication of a relationship between rate of roughness growth and application of friction modifier?
A
We have recently started to investigate this question.
1) The assumption of partial vs full slip should influence the shakedown limits.
2) The rate of damage accumulation vs the shakedown exceedence ratio is non-linear.
3) The assumption of Hertzian contact is not valid. The rate of contact stress reduction should be considered.
1) The shakedown map is based on the assumption that surface fatigue will occur if the highest surface shear stress (under the presumption of full slip) will exceed the yield limit in shear. This criterion could be applied also if partial slip is accounted for even if this increases the complexity of the criterion.
2) The damage accumulation may be related to a material Wöhler curve. This implies that the damage will be an exponential function of the shakedown exceedence rate. This explains the commonly observed trends of "epidemic" occurrences of RCF.
3) Hertzian contact relies on simplifications that may be more or less fulfilled. However, accounting for non-hertzian conditions is cumbersome, in particular since elastic analyses may result in unrealistically high contact pressures.
What is the frequency of the oscillations?
For example, at 100 km/h the frequency is 20 000 Hz. In the rail, the frequency obtained was equal to max 30 kHz.
1) Comment: Methodology first used/published by AAR (D.H. Stone et al) about 20 years ago should be a useful reference.
2) Cyclic softening (sigma_y > sigma_y_prime) is microstructure dependent, e.g. variation from coarse pearlite to fine pearlite. Hence bainite (concept grade) shows higher cyclic softening than pearlitic grades due to differences in carbide morphology.
3) Have you considered thermal effects?
1) The methodology is not new, but nobody in Europe applied it to railway wheel design.
2) Pearlitic grain size in carbon steels with the same chemical composition can be refined by a proper heat treatment working on temperature--time of austenizing.
3) No, until now, on wheel-set steel grades. We are working in this direction on hot work tools steels where temperature gradients are easier to reproduce (die casting process). A question is how to account for impulsive thermal effects due to sliding in low cycle fatigue experiments.
Given the material constants and coefficient of friction at the interfaces, how many independent parameters are there in your problem?
The following independent parameters are used in the model:
- relative clearance (R_1 - R_2/R_1)
- friction coefficient
- dimensionless normal and tangential forces and torque (dimensionalized on the material constants and radius).
Is the adjustment for the frequency response of the system based only on the multi-body calculations or have you used measured track response data for this?
The frequency response is based on the theoretical model, although in practice we use the data of the local station itself in an empirical way. This is necessary to account for the fact that we have stations on concrete mono-blocks, twin-blocks or wooden sleepers. This approach works quite well in practice. All stations work with exactly the same software, which is self-learning.
How sensitive is the GOTCHA system to the speed of the passing trains? Trains passing with higher speeds will produce higher dynamic forces.
Yes it is sensitive and therefore we only use measurements in a representative velocity range for the concerned rolling stock, e.g. rolling stock that operates normally at 40 m/s is measured at 30 to 40 m/s. At 10 m/s, you won't find any out-of-normal wheels anymore except for wheel-flats.
Can you detect subsurface and deep defects in the wheel material? If so, can you classify them and indicate maintenance intervals?
No, we cannot. We need defects that lead to dynamic excitation of the rail in order to measure rail deflections.
Observations presented by Mr Girsh showed that on the head hardened rails different wavelengths of corrugation develop compared to "regular" rails. This observations is in line with German tests in the early 1970's with different chemical compositions of rail steels. None of the presented corrugation mechanisms seem to allow for this. Can you comment?
For me it is extremely difficult to understand why a change in rail steel should affect the wavelength of the corrugation and I myself was surprised by the results shown by Mr Girsh. I should like to know more about how the "wavelength" was determined and what differences exist between the sites where the different wavelengths were measured before commenting further.
You did not mention continuously supported rails with respect to corrugation. What is your meaning about that?
Obviously they cannot get pin-pin corrugation, but they can suffer other types. Pin-pin corrugation has, however, also been seen -- I think because of insufficiently plane support.
On the BHP Billiton Iron Ore railway there has been some "squat"-type RCF damage in tangent track and shallow curves since the introduction of higher traction locomotives with steering bogies. In some locations this may be a result of automatic speed adjustment (ATP system), with transients in locomotive wheel creep. In other areas it may result from some bogie steering/wheel unloading effect on the exit from shallow curves. Wheel wear has also been higher with these new locos. Would you like to comment?
These are extremely interesting
observations. I have been told
by a maintenance contractor in the UK of some RCF damage that they
associate with high traction locos but this is the first report that I
know of possible damage associated with high traction locos on a heavy
haul railway, such as the BHPIO system. Although in neither
case
has the association between rail damage and high traction been
confirmed, it would in my opinion be surprising if rails were not
damaged by traction ratios of greater than 0.35 that some modern AC
locomotives in particular can maintain, with little wheel
slip.
If the damage were indeed caused by these locomotives, there are at least two explanations why you have observed this on the BHPIO system whereas no such damage has been reported from North America, where high traction locos are widely used. First, I know myself that the BHPIO system is extremely closely and well monitored by experts like yourself who have worked in this area for decades: you may simply have seen something whose significance has not been apparent to others elsewhere. A technical explanation is based on a claim that Steve Marich once made: that the 3-piece bogies and wheels on the BHPIO system are better maintained than is typical in North America. A possible result of this is that in North America (and elsewhere), the scores of wagons that follow a locomotive could "rub out" the damage caused by a loco, whereas on BHPIO (and generally also on passenger railways) the balance is swung the other way, so that loco damage prevails over wagon damage.
Another significant factor may be that curves on BHPIO are more modest than those on many North American railways, and similar indeed to those on higher speed railways.
Your experirence with increased wheel wear is contrary to that reported elsewhere with steering bogies on high traction locos. This may simply reflect differences in the bogie design. I would expect a forced-steering (rather than simply self-steering) bogie to give less flange contact even under traction, provided that the steering mechanism continued to work.
My understanding is that high traction locos are extremely attractive economically. Even if damage could not be controlled by harder rail steels, the cost of its removal by more frequent grinding of the affected areas may be a burden that railways such as BHPIO would be prepared to bear while railways elswhere could consider differential charging to account for the higher maintenance.
Rail grinding may also be a factor, given the closely-tuned nature of this system (one dominant traffic type and a fairly uniform distribution of wheel and rail profiles. The post grinding condition, including profile/facet detail & amount of material removed, sets the background/worst case with higher contact stress & cleaner/higher adhesion surface.
The steering on any bogie tends to create a hollow tread condition with the moderate curves in this system roughly equivalent to tangent & running on a narrow contact band. Under these conditions even a less than ideal steering will/should give a good result. However, 3 or so locos out of the fleet of AC's are problematic in terms of high & uneven wheel wear. That says the steering on specific axles could be negative with respect to a given hand of curve and certainly biased when it comes to tangent.
A hollow highly conformal wheel
tread is probably most upset by any
disturbance such as grinding. In the case of AC's the traction control
means you get the controlled squat type damage rather than a wheel
slip/burn.
In the end its high contact stress plus high surface shear that tends to be problematic in terms of squats & corrugations. Its perhaps an overstatement to classify the squats as a specific AC or high traction issue. The squat incidence at BHP is not widespread enough to justify such a generalisation. Quality control aspects regarding grinding and getting to know/managing the operating environment (especially every time it changes) are also important.
You mentioned the effect of traction force on curving performance, and the self--steering truck does not improve the situation. However, our experience in the development of powered trucks for commercial trains with self--steering truck from 1995 proved that it acts very well. The important thing is the attack angle in the leading wheel-set itself, not the differences between forced and self-steering trucks. How do you think?
In professor Suda's self-steering bogie, as far as I can understand from discussions, the wheel-set can rotate to some extent independently. Here the assumption made in our calculation of "normal" wheel-sets does not apply, and one might well find different results.
Although the written version of your paper makes no conclusion about the relationship between RCF cracks in UK vs powered and non-powered bogies, in your presentation you stated that powered bogies strongly contribute and non-powered bogies make little or no contribution to the development of RCF cracks in UK. Is my understanding of what you have said in your presentation correct?
Part I of the written paper concerns only forces on the track. Part II concerns damage. Both were presented, but only part I is written at present. Your understanding is correct for the vast majority of RCF that I have observed in the UK. It is not necessarily the case for RCF elsewhere. On heavy haul lines in particular, the direction of surface cracks is far more consistent with forces from non-powered axles ???
The traction ratio T/N you presented depends on the friction coefficient. Can you tell us the value of the friction coefficient used in the calculations presented?
The value used was 0.4. If a higher value had been chosen, I agree that even greater traction ratios would have arisen from some of the curving calculations.
Does heavy grinding affect the micro-structure?
It can, but the small martensite asperities that arise break off quickly under traffic and do not, as far as anyone knows, cause any damage. More lengthy sections of "blueing" are sometimes caused. These are ??? but again disappear quickly under traffic with no significant problem to the very best of my knowledge.
You imply that the problem of RCF in the UK may not yet be properly managed. Are you satisfied that Network Rail is going to be able to manage the problem?
Yes, but there has been, and still is, a great deal to learn about developing and implementing a preventive grinding strategy for a railway system that has been ground little in the past and where track maintenance generally has been run down for so long.
How to choose the rail profile when grinding to avoid two-point contact, in particular when running with flexible bogies where reversed steering could occur if the rail is ground without taking this into account?
The high rail profile, as emphasised in several publications, should but give so-called "two-point conformal" contact with the wheels. If the rails have been ground to give this profile for a fleet of old, poorly maintained bogies, it would not surprise me if there were problems when several new steering bogies are introduced. There is no easy answer as to how to select a rail profile to work in ??? ask if the wheels on the older trains should be allowed to remain in their current conditions.
5--7 MGT is a short interval. Has this anything to do with mixed traffic?
The 5--7 MGT interval is very much less than I myself have recommended. If lubrication were satisfactory and a sufficient relief given to the gauge shoulder area, I believe it is too short even for severe cases. I don't think mixed traffic per se would require a short grinding interval, but if powered axles are responsible for most damage, the criterion that will govern the grinding interval will be different from those determining the interval on heavy haul railways.
In the UK some rough relationship has been found between surface length of head checks and the crack depth. Has any relationship of this type been found in Germany?
It is very difficult to define any quantitative relationship between surface length and crack depth of "normal" head checks -- and therefore much more for this type. Our experience shows that only in the first stage of the development, such a relationship could be possible. In later stages, the surface crack length is influenced by wear, the profile, etc -- and we didn't find a useful relationship to the crack depth.
Why do you emphasize the distinction between "normal head check" and "horizontal crack"? The appearance of fractures clearly shows that fracture is driven by mode II exactly as for the "head checks".
It is not to emphasize any distinction. The horizontal crack is the result of a further damage process under given operational conditions. From the viewpoint of the practice, it is necessary to underline this. Such a damage stage has to be avoided by removing head checks in time.
We have experience of making clad rail with copper of 10 mm deep in about 1980. A 5 m rail was fabricated with explosive bonding to decrease noise by softening the contact spring. However, unfortunately as it was too expensive, the project was stopped at the trial stage of fabrication. What is the economical situation in your case?
The economic evaluation carried out by the end-users showed that for bottle-neck situations, for which the two-material rail is aimed, up to four times the cost of a standard rail is acceptable. This can be achieved.
What is the expected two-material rail life in MGT and what is the range of rail life (MGT) on a heavy haul line (Kiruna--Narvik or Kiruna--Luleå)?
At the Malmbanan test site, 900A rail had to be replaced due to severe head check damage at a 60 MGT interval. Since the two-material rail operates in elastic shakedown, replacement of this rail will not be be governed by head check damage, but by wear or high cycle fatigue. Wear resistance and fatigue properties have appeared to be excellent so far. On Malmbanan a life improvement by a factor of 6 is expected (as compared to the 900A rail).
1) Do you consider that wheel/rail contact temperature is one of the important factors that influences the potential coefficient of friction and adhesion between wheel and rail?
2) Have you observed any seizure effects at the coefficient of friction 0.8?
1) Yes, I do. I definitely agree with your opinion. However, it is not possible to measure it at this moment.
2) I checked the surface of the rail on which a friction coefficient of 0.8 was measured, and did not find any seizures.
As you have noted, one of the difficulties in implementing shakedown is to know what value of k (yield strength in shear) to use. There seems to be some debate on this. What, in your opinion, is the best practical way to determine k for a given application? If one uses Hv/b, do we use the hardness of the as-received steel or the fully strain hardened steel?
A first approximation is to use Hv in the middle of the strain hardened layer. Second approximation: Use Kapoor--Williams analysis of shakedown of a layered surface with layers of varying hardness.
Regarding the use of varied rail profile (tangent track) to distribute wheel tread hollowing:
1) Over what sections/lengths of the track are such profiles implemented?
2) Is there any difference in rail maintenance (grinding) requirements, e.g. grinding frequency?
1) In double track lines one
tangent track (TT1) profile can be used on "up" track. The other (TT2)
on the "down" track. In single Heavy Haul (HH) ines, half of
subdivisions
would use TT1, the other half TT2.
2) There is no difference in maintenance (grinding) of TT1 and TT2 rail profiles as both profiles are designed to yield about equal and low values of contact stresses as well as low values of effective conicity with respect to worn wheels.
A photo was shown of RCF cracks growing in the direction of traffic on the gauge and opposite to the direction of traffic on the field side. Is this not more characteristic for RCF on freight (and perhaps metro) lines than on passenger lines?
A
The photo in question was that of the high (outer) rail in 2.5° curve on a heavy haul line with ≈36 tonnes axle load. Gauge corner and field side bands of RCF cracks were generated by interacting with leading (_delta_r excess) and trailing (_delta_r defiency) axles of three piece bogies.
A comment: Scratch test shows similar behaviour to that encountered in abrasive wear processes, e.g., w/h hardness at surface and strain to failure are relevant (e.g., as discussed in the Lucchini paper).
1) Could more details of scratch tests be made available?
2) Was the steel prone to segregation and hence mixed bainitic/martensitic structures.
1) Scratch test details are to be provided.
2) No evidence of abnormal segregation and/or martensite formation was found.
What is your view regarding the reported sensitivity to stress corrosion cracking of bainitic steels?
There has been no evidence of SCC in either the laboratory alloys or the two heats of 136-RE rail produced for revenue service trials. However, after having been made aware of this issue with other bainitic steels, SCC tests will be performed on the J6??? rail prior to additional service test installations.
Have they measured the effect of J6 bainitic rail steel on normal pearlitic wheel steel, e.g. with twin-disc tests?
Wheel test disc information is not available.
1) Have you studied the correlation between dynamic fracture toughness test and charpy impact tests?
2) Do you plan to obtain dynamic toughness information with only charpy tests?
1) Research on the relation between dynamic fracture toughness tests and charpy impact tests has been extensively studied. However, at higher impact velocity (several m/s) or when brittle materials aew tested, fracture usually occurs at the very beginning of the impact event where dynamic effects are especially important. It has been demonstrated by caustic measurements that in this early stage the externally measured load is completely different from the actual dynamic crack tip loading. In other words, the load signal can no longer be evaluated by a quasi-static method.
2) I have no such plans. I am intersted in the difference of fracture toughness depending on the strain-rate of wheel-set materials for high speed trains. The static fracture toughness of wheel-set materials has been extensively studied by experiments, but the dynamic fracture toughness (K_ID) with respect to wheel-set material has not yet been studied enough, despite the possibility of fractures in curving lines caused by an impacting load. Furthermore, if the wheel-sets for a high-speed train were suddenly damaged by an impact load, the behaviour of the wheel-set would be determined not by the static load, but by the dynamic load.
Comment: At the beginning of the project it was thought that the reason that there is no well defined "critical crack length" at fracture could be due to the "locking" of parts of the long crack. From the simulations we do not see any such effects in the cases studied. However, these do not include very long cracks.
As the model is set up, it tends to produce wear in grooves across the wheel. How do you cope with this when you feed the worn profile back into the wheel.
Wear predictions so far have been generated by repeatedly running a vehicle over the same section of the track, which will mean that the wear is concentrated in the same region. When realistic mission profiles are run, the wear will be more evenly distributed.
The model of wear prediction essentially uses the results of wear simulation on twin-disc machines. However this test simulates rolling--longitudinally sliding conditions. But for flange--gauge face contact, it is very important to have results from rolling--lateral sliding conditions. Wear law and K coefficient may change considerably depending on P_gamma_-parameters (see paper in Wear by Zakharov, Komerovsky & Zkverov).
If the wear data is studied more closely, the "heavy" wear regime identified in the reference can be seen. In a paper I am here to present on mapping rail wear transitions, it is rather clear this "wear regime" can be seen in data from the literature generated using two-disc testing employing longitudinal slip and is not dependent on introducing lateral slip.
How do you consider non-dry wheel--rail contact conditions (weather conditions, man-made lubrication) in your wheel wear simulations? (I assume your wear tests in laboratory are only for dry conditions).
At present non-dry wheel--rail contact conditions are not considered, but this will be addressed in planned extensions of the work.
What is the form of the contact zone in your analysis? Elliptical or non-elliptical?
The contact is simplified to an ellipse in order to reduce computation times. Utilising Kalker's CONTACT would enable more realistic contacts to be considered, but would dramatically increase the calculation time.
What type of single-degree system is simulated? In the case of very high frequency circuit, the 3rd layer, such as water can deduce something called "condoneer??? effort". This effect can improve the conductivity. The "test" should be carried out with a wide range of frequencies from Hz to MHz.
Tests were run at 50 Hz. In future testing we could investigate this effect. I am not sure however at which range of frequencies UK track circuits are operated. It may differ considerably from that used in Japan.
Is it possible to use the pin-on-disc arrangement for testing the wheel--rail rolling contact or is it limited to the contact between gauge corner -- wheel flange?
Pin-on-disc testing is used to simulate the full-slip conditions seen in the flange contact.
Can you please explain why adhesion is lower in the middle of the day? Practical experience is that adhesion problems most often occur in the early morning due to mist or dew on the running surface.
It is so because there is not
so much water in the air then it is
during the night. The conditions are much softer and the hydrodynamic
regime is rather seldom xxx ? xxx
The system has been treated at 20 km/h even at a "large" curve radius (600 m). Would a higher speed at such radii have an influence on the results?
According to the experiments, speed has no influence on the lateral contact force. The main purpose of this test is to estimate the difference between "dry" and "friction modifier applied conditions", so test speed is kept constant. In the case that the running speed is balanced against cant, contact forces of the truck in the test stand are not significantly changed by running speed.
Is it recommended to apply friction modifier at initial friction coefficients of 0.3, or does the risk exist that the friction coefficient will become too low?
The answer will be presented by
Mr Makai for a practical case. In my stand tests the minimum value of
µ is 0.15. But the friction modifier has positive friction
characteristics, so slip or skid are decreased if they would occur.
1) How do you reallize curving conditions?
2) Can the rig test two-axle bogies?
3) Is the main purpose of the rig to study "wear" and not wheel-set dynamics?
1) Wheel--rail contact at curved rail conditions is realized by lateral displacement and angle of attack between the wheel and the rail, as well as different left/right vertical forces. Related values are taken from MBS-simulations.
2) Not yet, but after a future reconstruction.
3) The main purposes are fatigue, wear and lubrication, and testing of safety systems.
1) On the railway, rail does not get hot. Does the rail roller on the rig get hot and affect the wheel?
2) The regular out-of-roundness patterns seen on your wheel-set is typical of twin-disc tests. Are these patterns seen on wheels on tracks?
1) No, the system temperature
increase is maximum 5°C
2) Irregularities are seen, but not as the rail contact varies unlike on the rig.
Does the testing rig reproduce, in an adequate way, the dynamics occurring in reality or are there limitations or restrictions?
The macroscopic dynamics can be reproduced as good as a modern computer model can do. The high frequency structure dynamics of the wheelset and its bearings is reproduced quite well because of the use of a real bogie in the rig.
You showed that the sleeper spacing in particular influences the corrugation. What would be the effect of a contiously supported rail?
Sleeper spacing directly influences the pin-pin mode of vibration and therefore its growth. It's important to note that other modes of vibration may not be affected at all. A (perfect) continously supported rail will not have a pin--pin mode of vibration so there will be no corrugation growth associated with this mode. It is noted that this may not preclude the growth of other modes of vibration.
How well is the track's response at the pin-pin response represented by a mass--spring--damper? How greatly would any difference affect your results, especially with regard to difference in damping since you have emphasized the importance of damping on your results?
The growth of corrugation due to the feedback mechanism (at each wheel pass) ensures the system is self-excited at (or very close to) its natural resonances. Therefore in the present analysis it is assumed we only have to be concerned with matching the track's response at its resonances (modes). This enables the system behaviour to be well represented by modal analysis. The results obtained using the FE-model (which was benchmarked [8]) appear to confirm this (except when more than one mode dominates).
The effect of errors associated with the modal analysis may be quantified directly from the sensitivity analysis in table 1. For example, a 10% error in pin--pin modal damping estimation will be a -5% error in the corrugation growth parameter. It is prudent to assume that the modal parameter estimate (in most dynamical systems) will likely be one of the least accurate. For example, the modal damping, typically estimated from the sharpness of the resonant mode peak, may be difficult to estimate for the pin--pin mode since this mode is very lowly damped. It should also be noted that it may be deduced that errors due to the modal analysis approach have negligible effects on the results relating to the parameters k_0, Q/µP_0, µ, and I_R.
From "ordinary" fatigue design, it is well known that defects (such as notches, inclusions, etc) become more dangerous the higher the strength of the steel. Could this be an issue for the rail? Could it have an influence in welding of high-strength rails?
For rails, yes. And for early high strength rails it was a concern. However, there have been substantial improvements in rail steels (quality, cleanliness) such that current high strength (pearlite) rail steels possess relatively high fracture toughness levels. For wheels the situation is less so, although there are significant improvements over the last 8--10 years. For rail welds (FBW) no major concern as in high strength rails, particularily low alloy types. The hardness distribution is improved (narrower HA2 widths, higher fusion zone hardness); this reduces weld batter and impact loading. Situation slightly more complex for alumino-thermic welds, as using as-cast material where matching weld metal hardness to that of the rail invariably results in decreased toughness.
Comment: Distinguish between grinding intervention and respective number of grinding passes (depend on size and capacity of the machine); the number of passes in the tests may not be representative for routine grinding work.
The current grinding requirement (specified) is to remove min 0.2 mm and achieve specified profile to tolerance. Number of passes required is not specified -- left to the grinding contractor (and may have commercial implications). But larger machines favour -- achieve specified result in less cycles, and maybe better fit to profile. Possible benefit may come from installation of crack depth measurement equipment on rail grinders.
We have done a lot of measurements on the creep force -- creep rate relations. Frictions saturated at a creep rate of 2--3% at most. Your results show a rate of more than 6%. I believe there should be some existing "3rd body" in addition to the "natural 3rd body".
You think that "3rd body", assimilated at contaminants layers or wear debris, has an influence on the slope at the origin as many authors have attempted to explain. In our study, for many conditions tested, the big particles of natural 3rd body (type C) that you consider as "3rd body" are created after the saturation of maximum adhesion. In consequence, the only parameter which influences the slope of the adhesion--creep curve is the "shear threshold" of the subsurface, i.e. the natural 3rd body or, in other terms, its capability to flow in the tangential direction.
What is the mechanism behind the increase of the adhesion coefficient?
The evolution of the slope of the curve is correlated to the "shear threshold" of the natural 3rd body of the roller and the rail. Its activation in three rheological behaviours is governed by the rolling velocity, the creep and the environment of the contact.
Can you explain the dependency of the maximum adhesion on the rolling velocity? Are the results of the experiment in agreement with the results known from measurements on vehicles?
1) On site, the dependency is correlated to the dynamic effects of the vehicle. These effects increase proportionally with the rolling velocity. On the creep force--creep curve, the adhesion is measured in one direction, whereas the creep force at the contact has two directions. So the lateral effect affects the measure of adhesion. In laboratory, the dependency at a very low rolling velocity (<30 mm/s) has never been studied. I think the rolling velocity given corresponds to a 3rd body behaviour which depends on its natural "shear threshold" characteristics. The more the rolling velocity increases, the more the role of the natural 3rd body decreases.
2) The results obtained on site and in laboratory can't be in accordance because of the different dynamic effects (mechanisms) activated and the different parameters controlled in each case. As a consequence, the parameter of influence on creep force--creep curve obtained on site and in laboratory are compulsory different.
The natural 3rd body adhering to rail surface has been created by many previous conditions. The shear threshold of the 3rd body depends on the elastic properties of the rail material. So there is a thin, strongly adhesive, layer on the rail surface before the application of rolling--sliding conditions. The higher the shear resistance, the more difficult will be the removal of particles for a given creep rate in the micro-slip part of the curve. So the wear of the components is removed for higher creep rates when a thin layer of adhesive is present on the rail.
1) What are the criteria for evaluation of the influence of the 3rd body properties?
2) Is the influence of temperature, in particular on the creep velocities above the saturation point, considered in the analysis of the 3rd body properties behaviour and influence?
1) The criteria have been determined qualitatively. The low "shear threshold" corresponds to a nuance X. The high level corresponds to a nuance Y. In consequence you have the qualitative classification of the natural 3rd body properties.
2) The temperature created under rolling--sliding conditions gradually transforms the rail microstructure at a depth of a few millimeters. According to results presented by S Descartes at the conference, the layer of the natural 3rd body has a thickness of a few micrometers. Below this layer the microstructure has not changed. The mechanisms of 3rd body creation is the plastic flow of the subsurface until the formation of big particles adhering more or less at the surface.
1) What is the dominant frequency of the contact force?
2) How large is the unsprung mass?
1) The dominating frequencies of the contact force are the low frequencies, the sleeper passage frequency and the one induced by the corrugation of the rail.
2) The unsprung mass of an X2 wheel-set (the instrumented one used in the measurements) is approximately 1 205 kg and the axle load is approximately 11 750 kg.
Did you ever find the cause of the 3.1 m pitch depression on one rail at the Fåglavik site?
Our best guess is that they are caused by a wheel flat or a similar defect on another vehicle, for instance on a freight car.
What kind of micro-structure do you recieve after hardening to the 450 HB? Do you get any martensite?
No martensite. Just fine dispersed structure.
1) What type of defects did you observe in the wheel-sets after hardening them with plasma? (Similar to those occurring without plasma hardening? Occurring in the same zones?)
2) Did you ever think about using your system of hardening for the rails?
1) Correct technology leads to the improvement of all mechanical properties.
2) Plasma hardening leads to
the decrease of the friction
coefficient in contact with the rail.
What happens to the rail with respect to wear and RCF if the hardened wheels are used?
Plasma hardening provides a 2- to 3-fold decrease in the wear rates of the wheel-sets flanges.
There is a big difference between the hardness of the layer and bulk. How is the distribution of the hardness in the layers? Is there a danger for thermally affected cracks?
Base material ~280 HB. Hardened layer ~450 HB. Hardness distribution is soft.
Do you think that you could observe the same type of transformed microstructure without the mentioned heat effects?
The thermo-mechanical effect is one of three factors. Loose affility does resist fracture under the simultaneous effects of three factors.
The "perfect" rail also needs to have the correct transverse profile -- both after manufacturing and after grinding. This could be especially important for head-hardened rails, where low wear rates can cause high stresses to exist for longer -- raising the risk of rolling contact fatigue. Can you comment?
This is right. I know examples where customers were disappointed after use of head-hardened (HH) rails. These rails were delivered in the past without understanding the correlation mentioned above. So it is important to install HH rails milled in shape to the specific conditions of the railway. Only then is it possible to take advantage of the excellent properties of the harder material.
How does your controller handle the simulation when you suddenly go from a positive to a negative slope in the traction force/slip curve due to changing environmental conditions (for example at road crossings)?
Like in the reality, a sudden worsening of adhesion conditions leads to large longitudinal creep. The traction controller reacts on this new situation searching for an optimum tractive force. The algorithm checks the difference of the rotor torque in relation to the creep difference and stabilises the traction torque at a new working point.
Have you studied the evolution in time of the crystal degradation?
Yes, and I want to evolve this research. It is thought that the action of the rolling contact stress between the rail and the wheel can be described by examining the strain and the crystal orientation developed in the material.
Are they going to do further EBSP work at the other rail locations, where there is a tractive force (cf figure 2)?
In the present work, EBSP measurement was made only at the centre position where we thought there was no tractive force. We thought that at this position we could observe any phenomena which could be caused by the pure rolling contact fatigue without any disturbance from the turbulent plastic flow. In the next step of our work, we are going to do EBSP analysis at other positions than the centre in order to evaluate the influence of the tractive force and the resultant plastic flow on the further crystal orientation arrangement.
Comment: Align flange tips (assuming common m/c ??? profile)
Question: Did you consider the effect of bogie condition on the response?
We did examine the effect of bogie shear stiffness on vehicle stability. As the bogie is stiffened, the speed at which hunting starts increases and accelerations fall.
Does the stiffness of the track influence the simulation results?
No, because the results for new wheels are in good agreement. Only little influence is expected.
I noticed that the contact stress level has not been considered in the wheel profile optimization. Should it be an important factor to include?
Contact stress level is important for wear and rolling contact fatigue. It was not considered during the optimisation since it was assumed that in the case of trams the contact stresses are not high. It is indeed relevant for trains. Analysis of stresses can easily be included in the proposed optimisation procedure.
How can you explain the difference in results on head check development for nearly the same time period for the case of "9 months after grinding 2000" in comparison to the case of "10 months after grinding 2002"?
These are specific examples with particular conditions; the transverse profiles are different, as well as metal removed, so no general conclusion should be made from these specific recordings.
Are there any fundamental differences in grinding approaches that should be considered for freight vs high-speed lines?
In principle no, transverse target profiles are possibly different according to the wheel shape. High speed is more sensitive to correct equivalent conicity, but longitudinal profile must be smooth and transverse profile correct in each case.
Based on test results to date what appears to be the grinding rate (mm/MGT) necessary for keeping up with crack growth (as a function of curvature)?
At test sites: 18 MGT/year is suggested and metal removal ≈0.2 mm/year or 0.4 mm/2 years to be checked after the next grinding interventions.
When I started working at the rail administration 10 years ago, the first thing my manager told me was that if you neglect maintenance and lose track quality, you will never get it up to the same standards. The more I work with the issue, the more true I find it to be. Could you comment on the issue?
What your manager said is obviously true. Our experiences over the last decade have served to reinforce this by ménage??? Unfortunately, it is easy to save money in the short-term, at the risk of creating long-term problems.
Many of the accidents described have origins in poor practices and poor application of known knowledge. How can experts at conferences such as this better transfer knowledge to the people who need it to design/build/maintain railways?
This is a very important question. The dissimination of technical knowledge is extremely important. We should not be afraid to drive key managers from our research -- if necessary simplifying complexities in order to pass on key points. Sometimes as researchers we tend to underestimate the importance of dissemination to key audiences.
Can the system make matters worse if the friction coefficient is already critical (for example 0.1)?
Friction modifiers should not be applied to the condition of COF critical, such as oil lubricant supplied, because this will not make the situation better, but will keep the conditions as they were before the friction modifier was applied.
To what extent is the performance of friction modifiers sensitive to the application equipment? In particular, is it sensitive to the friction modifier film thickness?
The amount of friction modifier, i.e. the density of it, is an important parameter to control the friction. It is also important from an economical point of view.
A large reduction in the lateral/vertical ratio is observed after the application of HPF. As also the speed increases, is it related to a lower vertical load due to non-compensated acceleration?
The increase of the speed acts as an increase of lateral force of outer wheel and decreases the vertical force on the inner wheel. So the effect of a friction modifier is shown in this experiment.
You mentioned that the finite element (FE) models show differences from Hertz and CONTACT that are as large as 2 to 3 times. Could you please tell me under what general conditions this happens?
The same difference was found also using linear???-elastic versus elastic-plastic material in the FE-model. It happens on the rail corner where the radii of curvature were very small. On the rail head the difference was much lower. The elasto-plastic material has the same stiffness until the yielding point is reached. After that, the contact will be much softer. This results in enlargening of the contact area and in lowering the maximum pressure that affect the solution giving right length of sliding distance used in the Archard wear equation.
What is the expected implication of the contact force spectrum for the "multiple wheel" case on the rolling noise? Will there be a change in overall dB(A) level? Could the modified contact force spectrum have an effect when optimising (acoustically) the design of wheels?
We have published the effect on noise -- it has some small effects in the spectrum, but no difference is found in the overall dB(A) level. The effects are largest in the frequency range 400--1500 Hz, where the rail is the dominant source so that the effect of optimizing the wheel in this range has no significant effect on the total noise.
Could reflections account for the small variation of wavelength with vehicle speed?
Not for stiff pads where these observations are mostly made. For soft pads we expect a broader band of roughness to form from the multiple frequencies where growth is found.
We experience corrugation in Japan in the 100 ~ 210 Hz range. These are different from those in your research. Can you give some comments on the different formation mechanisms?
This is probably related to the
track anti-resonance in this frequency range. We have not considered
this range as wave reflections in the rail have no effect here.
Moreover we have looked at higher speeds where the higher
frequencies are important.
1) Which quantities did you use from your 3D-calculations to get the stress intensity factors?
2) Did you compare your stress intensity factors to any experimental results?
1) The displacements of the nodes where the external load is to be applied were used. These displacements are scaled with a constant including among other things the force applied to calculate the displacement.
2) No comparison is made to experiments. However, previously published papers on the influence function technique (by Åkesson et al) have shown excellent agreement to J-integral calculations.