N Scale Truck Reamers?

[C855B]

Here are the raw numbers, from eight years ago. The test was dropping rolling cars down a steep incline multiple times at velocities way beyond their normal use.



I'm open to other ideas, guys, so what conclusions would you draw?

[Maletrain]

The conclusion that I would draw from your data is that, if I care to run long trains, I need to test each car, no matter what trucks/wheels it came with, to make sure that it meets some standard I set for performance or cars on my test rig.  The smoothness of the wheel surface finish and the fit of the axle to the truck "bearing" are probably both important to the performance.  So, I would need to be able to change both to meet a stringent goal.

I have been playing with building a test rig that has a precisely adjustable incline so that I can accurately measure the angle at which a car (or truck) starts rolling.  Accurately weighing the car can then give me a scientifically based calculation of the starting force needed that others could relate to if they chose to build a similar system. I will make the apparatus adjustable for level, side-to-side, to eliminate the variable of side force from uncontrolled tilt to the side that might add different drag when set-up on different tables.  Unfortunately, my effort is currently stymied by loss of access to my shop.  But, this part of the project is straight-forward to do, when I get the chance to complete it.

Getting a scientifically relatable measure of rolling friction is a bit more difficult.  I would need to measure time or distance traveled along a section of track with the car (or truck) starting at a specific velocity.  Attaining and also measuring the beginning velocity accurately enough is something that I have not figured-out how to do conveniently.  I can think of several methods that cost significant money and effort - more than I think is appropriate.  Maybe I could make locomotive push the uncoupled car (or truck) at a known speed along the track until the loco hits a stop of some sort at the beginning of the measurement section, setting the car suddenly free at that point with the known velocity?

All of  this engineering would only gain the ability to state my data in standard scientific units. Unless others are going to make similar efforts, that would be of little value to me or anybody else.

So, personal, ad hoc methods are what most people will probably need to do.  Comparing the results of those in realistic ways may not be possible.  Apparatus variables like side-tilt, test rail smoothness, etc. may differ from person to person.  And, with 8 or 12 wheels per car (4 or 6 wheels per truck), individual wheel variables like tread smoothness, wheel wobble on the axle, wheel out-of-round, axle length variation, side frame tension variation on each axle, etc. get compounded into a bewildereing number of possible combinations that can make comparisons of measurements of things like "metal-vs-plastic" or " pizza cutter vs low profile" sensitive to a problem with just one of the wheels or axles in a test truck.

[DKS]

Perhaps one more factor (as if we need another) ought to be addressed: cleanliness. High mileage wheels almost always accumulate dirt, which degrades performance. In addition to putting to bed some old "plastic versus metal wheel" arguments, these testing procedures should also be able to accurately evaluate the effectiveness of various cleaning methods, since the results can be measured with relative precision and consistency. It will be interesting to see where this leads (assuming anyone takes up the challenge).

[Maletrain]

I agree that it is important to start with clean wheel treads and truck "journals".  The wheel surfaces seem to build-up black crud and the sideframes collect fuzzy stuff, both of which can clearly affect rolling properties.

But, just the act of cleaning might change the performance from "new" to " as cleaned" by changing the side frame tension when the side frames are pulled slightly apart to get the axles out to clean the sideframe pockets.  And, more than once, I have found an axle point sitting in an unintended divot in the side frame, rather than in the intended pocket.  Sometimes that seems to have been my fault, and sometimes it seems to have been the result of poor workmanship by a factory worker or a previous owner.

So, I can't assume that an as-purchased piece of rolling stock will roll freely enoungh based only on its nominal specifications (e.g., wheel material & profile), nor that it continues to roll freely over time, due to crud and fuzz build-up.  Even after it is cleaned, a quick test that it rolls as freely as expected seems warranted if long trains are in its service expectations.  So an easy and quick test seems appropriate as a routine measure, even though done with an apparatus for which the results cannot be related exactly to anybody else's results from their apparatus.

[C855B]

Interesting points about degradation with actual use, but my stats were for brand-new MTL cars with the factory air intact prior to the tests. I'm still interested in the "why" of the plastic axles worsening on subsequent runs.

My testing objective was only to compare metal vs. plastic, as the FVM wheelsets had just been released. The tests answered the core question, it's the unexpected results that I... uh... didn't expect.

[DKS]

...just the act of cleaning might change the performance from "new" to " as cleaned"...

Of that there is no doubt, and testing can help quantify this.

I'm still interested in the "why" of the plastic axles worsening on subsequent runs.

This is purely a guess, but IIRC the reason Delrin and related plastics are so slippery is that they exude a very thin oily film as they slowly and continuously outgas. It could be that quickly-repeated tests temporarily deplete this film. Again, this is merely speculation.

The tests answered the core question, it's the unexpected results that I... uh... didn't expect.

And therein lies one of the thrills of scientific discovery.
 

[narrowminded]

Perhaps one more factor (as if we need another) ought to be addressed: cleanliness. High mileage wheels almost always accumulate dirt, which degrades performance. In addition to putting to bed some old "plastic versus metal wheel" arguments, these testing procedures should also be able to accurately evaluate the effectiveness of various cleaning methods, since the results can be measured with relative precision and consistency. It will be interesting to see where this leads (assuming anyone takes up the challenge).

And this is an example from above where I was testing pizza cutters and inadvertently tested the dirty wheel effect too.

Quoted from my post:
"I had some pizza cutters in the junk box so assembled a couple of trucks with them and just ran them.  Initially one ran eight cycles and one ran nine.  A careful inspection (as well as the sound they made) showed that they had the typical gunk buildup found on dirty wheels so I took them back out, cleaned them, and sure enough, 10-11 cycles for both trucks.  No change from the later RP-25 wheel profile.  Very consistent running."

What I've already seen that is satisfactory to me in a clear and quantifiable way with this simple rig is that:

1) Metal wheels do perform measurably better than plastic.
2) Adding the weight of a second truck made very little change in results from unweighted for either material.
3) Dirty wheels effect the roll resistance measurably.
4) Clean pizza cutters had no different roll resistance than low profile.

What I haven't seen evidence of in the few simple tests I've done is the results of C855B's tests as they apply to the plastic wheels.  The metal wheels tested superior in all tests, his and mine, and both were consistent.  What doesn't make sense to me from my experience vs: his is the diminishing performance with the plastic wheels, pizza cutters or low profile.  I didn't see any change in test runs more than five but less than fifty.  He saw a diminishing performance change over just five runs.

As I explained previously, I don't think it's heat as there just isn't much to be generated in the basic design, almost imperceptible (unless the speed and duration were beyond what I can imagine).   I'm not doubting the results but wonder what else may have effected them.  Dirt would be one if they were older wheels or frames.  Damage to the axle points while putting the wheels in and out of the frames might be a factor but then, why diminishing?   I wonder if there is any static electricity built up that has some direct effect or maybe attracting dust?  I don't know and will leave it to others to try to rig some tests and figure that one out.

And I didn't think the pizza cutters roll condition would be any different than the low profile based on visual inspection of the engagement with the rail.  Performance tests confirmed that for me and with no surprises.

[narrowminded]

Interesting points about degradation with actual use, but my stats were for brand-new MTL cars with the factory air intact prior to the tests. I'm still interested in the "why" of the plastic axles worsening on subsequent runs.

The ones I used were of questionable age and use but I think minimal based on visual.  Except the pizza cutters.  They had a nice crud build up.   Four different frames were used and for confirmation of the frame's effect I did various swaps between axles and frames.  The results were basically unchanged. 

I also did this three different times now putting the rig, trucks, and wheelsets away, then getting them out again and reassembling for another look.  There has been no change in results in any of the duplicated tests.  The plastic wheels, cleaned pizza cutter or low profile, keep doing the same thing no matter how many times I've rolled them or which configuration of truck to axle was assembled.  The metal wheels also repeated but at their measurably higher performance level.

[Maletrain]

The results show that all 5 of the plastic wheelsets decreased in rolling ability rather monotonically over 5 successive test, while the 2 metal wheelsets appear to be unaltered in rolling ability after the same number of tests.  But, there was a wide difference in the rolling ability of the 2 metal wheelsets, and also in the 5 plastic wheelsets.  They seem to have initially spanned about the same range.  So, it is hard to say that the plastic wheels clearly out-performed the metal wheels, or that they would not have quickly degraded to worse than the metal wheelsets in normal service.  In other words, your test have raised yet another variable to consider: how quickly does rolling resistance increase with service time for various wheel and truck material (and shape?) parameters?

The differences in the plastic wheels might be that they pick-up crud faster than metal wheels.  Where the crud came from might have been the plastic wheel material, itself, transferring to the rails and back to wheels.  Or, it might have been crud the factory left on one or more wheel sets that then transferred to others during the tests.  Or, maybe you had oil (WD-40?) on the test track that affected the plastic but not the metal wheel treads?

But, your data for the plastic wheels show that they degraded over 5 test rolls by 20%, 6% and 8% for the pizza cutters and by 3% and 7% for the low profiles, while the worst performers went from 81% of the performance of the best performers at the beginning to 64% after 5 tests.  So, one of the plastic wheels sets degraded much faster than the rest, but it was one in the middle of the performance range to start with and was still better than the plastic worst performer at the end of the 5 tests. 

I think this all points to the suspicion that there are individual problems with wheelsets that will tend to be larger than the differences between types of wheels.  And, your data seems to demonstrate that those individual differences cause different variations over time that are important, as well.

Even if we manage to get data that unequivocally showes that on average, metal wheels or plastic wheels roll better to  begin with or degrade slower in-service, we will only learn which is best to start with when we buy wheelsets.  Because the large variations among wheelsets of the same type seem to be greater than the differences between the averagesof the types, we will still need to do our own testing to weed out the bad ones, or for really long trains, pick out the best ones.

[Maletrain]

Narrowminded's test results with old plastic wheelsets not rolling as well as metal wheels do not seem inconsistent with C855B's tests with new plastic wheelsets that show them rapidly decreasing in rolling ability with "use."  Perhaps if C855B continued his tests to the point the his plastic and metal wheels were as "used" as Narrowminded's wheels, the plastic wheelset performance would have degraded enough to match Narrowminded's results.

[C855B]

One difference to consider in my tests is I had an 'S' curve after the incline to test flange loading. I don't think it would have influenced loading on the axle points, but it could have. The launch ramp was quite steep - thinking back, it must have been close to a 50% grade. No worse than the foam test arc, but the run was longer and the relatively high speed sustained on the level until the friction won.

As to the wide variances, I will chalk that up to inconsistencies in the sideframe pressures. I mentioned before the concerns about the variability induced by removal and replacement of wheelsets. As to the 5th and 6th lines in the chart, one of those series has to be an outlier reflecting... something. As to the plastic shedding enough in a 10-foot run to influence subsequent results? I doubt that.

As said, there are a lot of other factors that could be considered here, but for the conditions of my test, the dirt/use issues shouldn't be a factor. Track was not oiled, it happened to be new Unitrack.

Again, I was testing for plastic vs. metal, the other results are incidental curiosities, especially the diminishing roll-out. It's not the absolute numbers I'm puzzled about, as that's going to be, to me, normal product and technique variability - it's that every plastic wheelset experienced the degradation versus none for metal.

[narrowminded]

One difference to consider in my tests is I had an 'S' curve after the incline to test flange loading.

Again, I was testing for plastic vs. metal, the other results are incidental curiosities, especially the diminishing roll-out. It's not the absolute numbers I'm puzzled about, as that's going to be, to me, normal product and technique variability - it's that every plastic wheelset experienced the degradation versus none for metal.

That was baffling to me, too.  But something you just mentioned, the "S" curve, adds a theoretical possibility for me. 

The plastic axle set may have more traction than the metal set.  The metal vs: plastic tests suggest that's true.  And if so, then going through the curves with two wheels locked on a live axle, travelling on two different radii, introduces a binding effect requiring at least one of the wheels to slip.  Additional traction/ friction means more energy consumed overcoming it, more slowing.  But why the diminishing effect?  Still don't know but maybe some scuffing that improves the "traction" and aggravates the problem.

Something that could be useful from this thought exercise is making the case even stronger for metal wheels.  In a long train, what are the chances that you won't have at least one section of your train pulling through a curve and maybe more than one.  Well there's a further worsening effect with plastic caused by that bind... if it exists.  I'll bet there's something there and when that effect is added to the improvement demonstrated on a simple straight run with metal wheels it becomes very apparent that a REAL benefit can be realized with metal wheels if you're pulling long trains (or just trying to pull more cars than you can now).  Might be some measurable help there.  And somebody with a multi-level helix probably already knows this even if not knowing why... if it's even true.    And I really need to quit this now.

[Maletrain]

When you say "outlier," that implies that something is wrong with the way the data was taken, or maybe that the sample was not "proper" in some way, either suggesting that the related data should be removed from the analysis. 

For your data, I don't think so.  The way the line 5 plastic wheel data trended downward over the 5 tests does not seem to indicate that it differed from the other plastic wheel tests in any significant way.  The most noticeably different perfomer was line 1, where the rolling degradation was much greater with successive tests than any of the others.  But, that too, was a consistent trend, saying more about the individual wheels than the test.

People who use statistics in search of "knowledge" too often focus on the averages and not the variability, thereby kidding themselves into having more certainty about their conclusions than is actually warranted by the data used to reach them.  That tendency too often leads to bad surprises in the future, when reality showes the conclusions were wrong or unreliable for individual cases.

[peteski]

People who use statistics in search of "knowledge" too often focus on the averages and not the variability, thereby kidding themselves into having more certainty about their conclusions than is actually warranted by the data used to reach them.  That tendency too often leads to bad surprises in the future, when reality showes the conclusions were wrong or unreliable for individual cases.

Isn't there a saying that states "lies, damned lies, and statistics"? 

[Maletrain]

Isn't there a saying that states "lies, damned lies, and statistics"? 

Yes, that's the quote, and true enough.  But, there is a difference between lying to others using improper statistical techniques and fooling oneself by ignoring or down-playing the uncertainty in the conclusions due to the variability in the data.  The first is intentionally deceptive, while the second in usually unintentional, or at least unconsious.  Data analysts who really want to objectively figure something out are still prone to thinking they have done so, even if the data is not really very conclusive.