Edward Thompson: Wartime C.M.E. Discussion

I think the answers would be 'yes' (although it would need good data on the then-current geometry of the track on the Edinburgh to Aberdeen line; no idea if that still exists) and 'yes' (an additional late change to the axle design, per the design study document, was to increase the diameter of the outer portion of the axle to 10", which will of course significantly increase its load-handling capacity; in addition, the news item which I linked to above indicates that the trust regularly inspects their axles with ultrasonic tests, so there is an additional margin of safety there).

An interesting question, but I suspect that the details of the circumstances (e.g. at the time the P2's were failing, they were in daily use on the curvy Edinburgh to Aberdeen line) are sufficiently different that they may not be usefully comparable. Also, apparently all the failures happened "at low speed when the locomotives were accelerating hard", which makes it very likely that the increased torque of the P2 was a major factor in the failures.

Well, you can't get track spreading without the application of sideways force by the wheels! I wonder if there are historical records of track spreading on the Edinburgh to Aberdeen line during the period the P2's were using it? (Although of course there was almost certainly other traffic using the line, which could have contributed.) That would be a telling indicator, though. But such a sideways force would of course add to the other stresses on the axle.

Which brings another thought to mind. The increased torque at starting would be in a different direction (rotational, inside the shaft) from the forces of i) the weight of the engine (transverse, but at a right angle to the rotational axis of the axle, trying to rotate the end of the axle up) and ii) the proposed sideways force exerted by the track (transverse, but parallel to the rotational axis of the axle). (Alas, that apparently would not leave any diagnostic indications if the failed piece.) Those different directions, along with the rotation (with respect to the axle) of those two forces, would make any modelling that much more complicated, especially for study of crack formation and propagation. Crack formation and propagation (two separate phases of activity) are tremendously complicated, and modelling them is still a nightmare, even without such a dynamic load situation. E.g. the time-varying load directions might have more of an influence on the propagation phase than on the initial formation.

Just to cap it off: usually, steel can be considered an isotropic material (the same physical properties in all directions), and it mostly is (compared to, say, composites, like carbon fibre). However a metallurgist will tell you that when you want really detailed, accurate models, it isn't - there's grain structure which depends on the manufacturing method (purely cast, or also forged, etc), which usually has differing effects in different directions. So really accurate modelling of the original failures might need details on how the initial axles were fabricated.

Not that I'm at all concerned, though; between use on other, straighter, lines, and the design changes (especially the increase in diameter) I suspect the problem is unlikely to recur.

Noel

 

Well, they only mention the starting torque; they make have looked at other forces, and found them un-important. And the fact that the failures all happened on starting is 'interesting'.

Along with the use of data accumulations, not anecdotes! ('The plural of anecdote is not data'.)

Noel

 

This amateur reader seems to feel that most the discussion is centred on sideways motion BUT given that Fife was a major coalfield in the 1930s, I wonder if track movement in a vertical plane could be equally a source of stress. I accept the curvaceous nature of the line - especially in Fife - but I wonder what consequence mine workings would have on both track stability and the consequent source of stress on a lengthy (and weighty) locomotive.

 

A few points to bear in mind.

The P2 axle failures have clear evidence of stress raiser/sharp corners. This has been resolved by the recent P2 analysis.

Track spreading looks like a red herring given how many WDs were in Ferryhill over the years.

Steel and controlling the grain structure have moved on considerably from the 1930's.

just read back on this "sideways load" discussion.

FE like so many things is about what you put in to it.

This is an axisymmetric problem and we are talking about effective bending.

The underlying requirement is to numerate the predominant, primary, force. From there calculate the principal stress.

If you take the axle as a beam and support that beam where the axleboxes are, you will see the supports are sufficiently close to the application of the load so as to deliver small/negligible bending moments. Put another way, the wheels are too close to the boxes to deliver significant bending.

Shear forces thought the cross section will exist, but eyeballing it, 11 tonnes (two supports remember) over 9-5/8 axle, it's bending of low thousands of N/m. Not a concern. I assumed solid axle here.

It then becomes arguable that the more concerning issue is the torque applied if the axle is not purely circular and sports a keyway. Which it did. Stresses in the corner of that keyway likely exceeded what you would tolerate on the S/N curve for cyclical loading.

What is being said is true, bending and torque are the combined loading's on the axle. That is factual. But the bending component is far less important than the torque in this particular example.

 

If the track on a curve doesn't give, then something else has to.

 

A WD is a totally different engine designed for different duties. It just happens to have 8 driving wheels. There is no guarantee that it applies forces to the track in the same way a P2 does. It's a bit like saying a class 37 has the same number of wheels as a P2, and that doesn't spread the track.

Tim

 

But the track on railways does give. Its not totally fixed and does deflect somewhat given the forces it encounters.

Except that it inst and I did not say that. Nor would I. A P2 isn't articulated for a start. It will apply different forces, but if the argument being pushed is that 8 coupled locos spread the track....which incidentally is not my argument. I think that one is a red herring.

 

There seems to be an assumption here that the P2 was uniquely long in its fixed coupled wheelbase in British practice. Obviously most ten-coupled engines were longer but had flangeless centre coupled axles. There is one other class of interest though - a fairly small but long-lived and apparently successful class - the GWR 4700s. These actually had a fixed wheelbase 6" longer than a P2, and the wheels are not that much smaller at 5'8". Same 6'6" spacing between the first three axles, and 7'0" to the trailing axle. I don't believe (but am open to correction) that they had any provision for side-play on the coupled axles. Given that they were also intended to spend time going in and out of freight yards, well known for their excellent trackwork, I would think that if track spreading was going to be a problem they would also have developed a reputation. 8-coupled chassis also have a slight advantage over 6 and (fully flanged) 10-coupled, for a given total fixed wheelbase, because there isn't an axle in the middle, so they can cut the corner ever so slightly more.

 

Any spreading would be a symptom, not a cause.

Per the design study, most of the failures happened "immediately behind the wheel". That doesn't prove anything, but it certainly seems curious - as does the choice of the original designers to increase the diameter in that region. I suppose the latter could have been to allow the use of a larger bearing, but a cylindrical shim could have done the same.

Without doing a lot of calculations, I don't know if that assertion is correct or not.

I go back to the observation about the observed location of the failures. Of course, there are also the observations that "all these failures took place at low speed when the locomotives were accelerating hard", which argues for a key role for the starting torque. And given that the "design of the P2 crank axle is essentially the same as that for the contemporary A3 Pacifics which were not prone to axle failure", one has to reinforce that observation.

But the consistency of the failure location (and possibly also the failure plane) do argue for bending forces on the axle ends having a key role too. Alas, it would take a lot of calculations to know for sure.

PS: To put it another way, the fracture, to be consistent like that, had to have happened at the point of maximum stress. But the torque would effectively identical all long the axle. Ergo, there had to be a source of additional stress at that point.

Noel

 

Locomotives of the GWR Part 9 says that the two inner coupled axle wheels had thinner flanges and the coupling rod bushes had spherical seatings.

 

The argument should be did the P2 spread the track? Then collect data that can either prove or disprove that hypothesis. I wouldn't put the existence of another loco nearby with an 8 coupled wheel base in either of those two categories. Without further investigation there isn't enough evidence on its own to conclusively use that information to prove or disprove the hypothesis. Correlation does not imply causation after all.

The Vampire model appears to show very high lateral loading on the third driven axle and the leading pony truck. This is quite a contrast to the Pacific which has very even lateral loading on all axles. Of course to know if that possibility spread the track you'd need to know what the condition of the track was like.

Tim

 

Well indeed. We have no evidence to show that P2s spread the track any more than any other machine abs so I don’t see why others are proposing such a theory.

 

Its important to note that the key way reduces the area, and subsequently increases the stress for a given load condition. That point of maximum stress is that sharp corner in the key way, which also happens to be the smallest cross sectional area of axle.

That sharp corner of that key way makes a crack relatively easy to initiate under a torque load. Or indeed a bending load. But even doing a rough calc, I have my doubts that bending is more than a secondary contributor. Unless it’s the same axle every time?

Once that crack has been initiated, the cyclical loading-unloading, the hammer blow, the piston thrust, whatever driver, makes that surface succumb to fatigue fracture advancing through the cross section.

One day you find yourself at Stonehaven, the regulator is pulled, the loads go on to the axle and there just isn’t enough cross sectional area left, the stresses exceed the ultimate strength of the material and hey presto, one liberated wheel.

There is no doubt that this problem has a dynamic load. There are torque, bending, and cyclical loads at play. Which one is the ultimate culprit belies the point. What really matters is the axle is inherently flawed for the load condition presented.

As I see it, given that axle loadings on the P2 aren’t any higher than the Pacific’s, which had similar axles with similar key ways, the differentiating aspect is the adhesive capability. Torque in the axle isn’t constant as there are multiple pistons exerting multiple torque loads with each full revolution. Torque will also be cyclically applied and removed and in some instances will be doubling up as more than one piston pushes at the same time. Maximum torque values are bounded, but the loading condition through the axle is cyclical in relation to piston/crank forces.

To my eyes it’s a crack initiation then a fast fatigue fracture.

Out of interest, which axle was it that failed? Was it a leading driver? Or were there others? Was it always the same axle?

 

This discussion was about 'why did P2 crank axles fail'. I can't see how a spread track could cause an axle failure, so all this focus on spread track is ... basically irrelevant.

Ah, no. The failures were just inside the wheel; the axle is 9-5/8" in diameter there, and only 8-1/4" in diameter next to the crank web.

The news page says they were the crank axle. From the images there, that seems to be the leading axle (which makes sense; if not, the rod from the center cylinder would interfere with any passive axle(s) between the center cylinder and the crank axle. (Interestingly, all 4 wheels were coupled together; see the diagrams. The main rod runs to the second wheel.)

Noel

 

It's a charge I've heard made against several designs over the years, but having seen nothing convincing to substantiate just about any of them, remain sceptical

One of the few exceptions were a brace of 4-8-0Ts, designed by E.A.Watson, built for service in Dublin's Kingsbridge goods yard, the only other duty mentioned was banking goods trains between Kingsbridge and Inchicore, plus a nocturnal goods banking diagram to Clondalkin.

A short lived design by any standards, let alone those in Ireland, quite what prompted such a leviathan for this task mystifies me. A year before withdrawal, No.900 had it's rear coupling rods removed, turning it from an ungainly 4-8-0T into an ungainly 4-6-2T and losing 15t of adhesive weight in the process. Given No.900s propensity for derailing and generally being not very nice to the many tight curves in the yard, the only thing odder than it's initial introduction, in 1915, was Watson's successor, Bazin, building a second example, in 1924.

No.900 was withdrawn in 1928, No.901 went in 1931, having seen only a paltry 7 years service. They were the only 8-coupled locos ever built for the Irish 5'-3" gauge. Why on earth the GS&W didn't consider a Garratt, when one ruddy awful machine proved insufficient, I can't fathom. Certainly a Std Gauge 0-4-0+0-4-0 of comparable power and weight was developed for a more demanding role in the S.Wales steel industry in 1924 (and formed the template for future small industrial Garratts). Maybe it was the unfamiliar idea of bendy engines, or the Garratt's much smaller coupled wheels (3'-4" against the 4'-61/2" of the 900s) which coloured GS&W thinking, or maybe Watson's 1921 departure to (wait for it....) Beyer Peacock.

To be fair to Watson, the requirement these machines supposedly met had previously defeated no less an engineer than his predecessor, Richard Maunsell, though what that gentleman learned in the process was later put to good use in the Southern's Z class.

No.900 at Kingsbridge in it's last year of service (rear coupling rods removed)


[Image courtesy transportsofdelight,smugmug.com]

 

I'm confused. The crank axle on the P2 is the second driving axle onto which all three cylinders drive. It's a unified drive machine. As I understand it all failures occurred on this axle.

 

Most British 8-coupled designs were freight engines, typically operating at slow speeds where some of the forces and stresses would be lighter than for engines at express speeds. As well as having small wheels, most of these designs had quite short wheelbases. For example, the WD 2-8-0 had a coupled wheelbase of 16ft 3in, shorter than that of a Midland/LMS 4F (16ft 6in) or LNER J39 0-6-0 (17ft 0in). So even without the benefit of a Krauss bogie, they should have gone round corners fairly easily.

I believe all of Gresley's 3-cylinder engines had unified drive onto the 2nd coupled axle, except for the B17 (divided drive) and D49 (front axle drive).

 

Ah, you're right. This picture confused me; in part because it looked like the rod from the crank to the center cylinder would foul a preceding plain axle. But after some looking, I found this image which does indeed show that the crank axle is the second axle. [I think we can assume that the Trust's CAD image is correct! :-]

Noel

 

The exception (at least to an extent) were Churchward's 47xx. Of course, with just the two outside cylinders, but with 5'-8" dia drivers giving them a fair turn of speed, they were latterly in demand for heavy summer passenger trains.

I'm not familiar with the GW's definition of fast fitted freight, but wonder whether the wheel balancing, presumably optimised for their designed operating speed range, caused any undue issues when pushed harder on passenger turns, though from personal recollection of such workings from Paddington to Weston-Super-Mare, in the charge of Class 52 Westerns by the mid 1960s, such trains weren't exactly in 'Cheltenham Flyer' territory, speed wise.