Tangmere

I'm not sure I understand what this is getting at - is it i) that were such a rule to be put into place, it would remove most of the steam locomotives which are currently allowed on the national network, or ii) with so many locomotives of that design about, experience has proved that it's not that problematic a design. (Or both of the above, perhaps together with others I've not though of?)

If you meant the second, I agree that that's a good point - but I'm not sure how much water that holds these days. My sense is that in some quarters these days people are unwilling to run risks, and are willing to spend money on very incremental safety improvements. (I don't particularly agree or disagree with that thinking, but that's just the way it seems to be now.)

As others have pointed out, had any injuries or deaths resulted, there probably would have been more pressure to 'do something to make sure this cannot happen again'.

A very good point. And one can also point out that there are other major motion links which could produce almost identical results were they to come lose (e.g. coupling rods), so even doing something about the connecting rods would not remove the threat of this kind of failure.

Noel

 

Noel - I was making reference to the number of locos we would loose from the mainline if locos with single slidebars were banned. I do agree withyour second point as whilst the statistics show they are reliable if they do fail it can be with dramatic consequences. Before retiring I was involved in "what if" planning. There was one scenario, which unfortunately I cant go into, which was so extreemly remote as to be discountable but the consequences were severe. In fact the consequences were so severe that real money was spent planning how the hell we get out of this mess. That is where we could be if Tangmere had derailed

 

Noting this BRS 102 / 6 is the old BS Report 24 Number 8 Class D specification which was later regraded as BS 970 / EN8; I have seen references to a 1955 spec but latest details refer to a 1995 spec for this but even this has since been redefined as noted at :

http://www4.hcmut.edu.vn/~dantn/Matter/Strength_st.htm#equivalents

Referring back to the original RAIB report, it documents the footplate crew hearing a loud bang; could that be the assembly disintegrating in an explosion hence the difficulty in finding any parts ?

 

Well there is 3 other things that may have contributed to the failure other than the nut 1 the piston was loose on the piston rod causing vibration ,2 the hole for the cotter was 90 degrees out of position it should be vertical not horizontal and 3 the taper between the gudgeon pin and cross head if in good good condition would require a lot of force to separate it.

 

The damage to the gudgeon pin is interesting in that only the outer few threads are worn away evenly all round and the key in the back taper is worn significantly more over the last quarter or less. This suggests that the gudgeon pin was partially out of position for a relatively long time. If the nut had been unwinding slowly after the cotter had been lost, it should have protected the end of the thread until it fell off, whereupon the gudgeon pin would have fallen out almost immediately. There wouldn't have been time for the end threads to have been so evenly worn down, and the key would have been worn evenly as the nut allowed the pin to move slowly inwards. Conversly if the nut had split and fallen off, the gudgeon pin would have partially retracted into the crosshead and stayed in that position until the cotter was worn through and fell out. That said, I struggle to imagine quite what loading conditions would cause the nut to spilt. A pity there are no photos of the damage to the crosshead.

 

So I went back and carefully re-read the report, and then asked my SO (who is a structural engineer in the aerospace industry, and thus has extensive professional expertise in this technical area) to read the report so I could ask her about the bits I wasn't 100% sure about, to confirm my tentative understanding.

The thing is that the major problem was not that they'd used the castellated nut, but in the design of their 'home-made' cotter pin. (And if you already understood that, apologies - it wasn't quite clear from the wording.) There were two issues there:

First, the cut (to form the two legs which were bent out) only went roughly as deep as surface of the nut, which had two bad consequences . Most importantly, that meant that the bend at the base of the 'legs' was close to the end of the cut, locating the bends at a stress concentration (see paragraph 65, second part). Also, the fact that the part of the cotter inside the pin is a single piece of metal meant that the cotter was free to move slightly within the pin (paragraph 65, first part). To put it more simply, the second issue meant that the cotter could move about a bit (making it more likely to cause crack propagation over movement cycles); the first meant that the legs were more likely to break off.

Second, the head of the cotter (the bit that retains it from the other end) was formed with a right-angle cut - again, a stress concentrator at the 'tip' of that right-angle being likely to serve as the initiation point for a crack (para. 66). (Right angles are always bad news - the ones at the corners of the Comet windows were what caused the famous Comet crashes of the 1950s.)

These issues would have been the same whether the nut was a plain nut or a castellated nut, and were likely the cause of the failure.


Now, it turns out that the plain nut actually is likely slightly superior, for two reasons: one fixable, one not. The one that can be fixed is that if the slots of the castellation are cut with straight bottoms, again there's a possible stress concentration - but that can be avoided if the bottoms are radiused. (Small chance of a crack there, but still...)

The real issue with the castellated nut is that ideally you'd like the nut tight to something, so it doesn't move around over time. To preload it against the nut face would take about 2 foot-tons of torque (see para. 48) - one heck of a lot of torque! However, it can be preloaded against the cotter pin with a lot less - which was apparently BR practise 'back in the day'. (See para. 58) One obviously can't do that with a castellated nut.

So in that minor aspect, the plain nut is better - but they are quite close. The real issue is the cotter pin.


Anway, sorry that went on for so long - hope it was of some interest. (My SO also had some comments about the materials issue - I can pass them along too if there's any interest.)

Noel

 

The details of the cotter are only relevant if the nut itself becomes loose and started trying to force the legs and top off the cotter as it rotated. Whilst the nut remains tight the cotter could be made of chewing gum and not affect the situation.

So, if the nut was tightened with flogging spanner and hammer to pull the pin into the two tapers, why did the nut come loose? (assuming the cotter fell apart before the nut came off - see my post above)

Furthermore, good practice is to never back a nut off once it is tight, either to fit cotters or to tighten them. For a castelated nut, the whole joint is resplit and the nut skimmed so that when retightened the cotter will fit the castellation; it should also be a tight fit in the slot. For a plain nut the cotter is remade to be a driven fit against the nut.

 

As with most failures or accidents (in general, not just railway related), it is rarely the result of one thing going wrong, its a chain of events which happen.

A interesting report though, with lessons to be learnt across the whole heritage movement. These machines use big heavy components which can cause devastating problems if they go wrong.

 

True, and if the failure had happened away from the third rail, that added link in the chain could have been devastating. That doesn't mean it isn't worth trying to find the root cause at the start of the chain.

 

I'm not a MechE myself (although I have spent 30+ years taking a significant interest in my SO's professional activities, and I am an engineer, so I do know a fair amount about this subject), so take all this with a tiny grain of salt, but...

My understanding is that that is correct.

Ah. See paragraphs 47-48 of the report. I will leave off a long discourse on the role of elastic deformation of a bolt in holding a nut on, and echo the report in saying that a number of different processes can cause the pre-load in a tightened nut to be lost (allowing it to rotate freely under the influence of vibration, etc).

One that the report particularly calls out (see para. 48) is embedding loss, which is (over time) plastic (i.e. non-reversible) deformation of the tiny roughnesses on the surface of the nut face, etc. This causes a reduction in the pre-load - and if the initial pre-load was small enough, complete loss of the pre-load.

The report says that to prevent complete loss of pre-load via this mechanism in this particular (massive) joint, one would have needed to apply a pre-load of 3000 to 4400 lbf-ft; in other words, with a 2' spanner, you'd need a force of roughly one ton on the end of it to produce enough torque to produce that level of pre-load. 'Now give it a good one with the maul, Ralph' isn't going to do that!

Right; I was a bit puzzled by the reported practise (para. 58) of backing off the nut - I thought, 'surely that would leave the pin loose, or allow it to become loose easily?' The only answer I can find seems to be in para. 46: "When the locomotives were built, the gudgeon pins were required to be lapped into the crosshead .. which would make the two components lock together, requiring some force to separate them. Dismantling a small end assembly ... normally requires the use of a jack or punch bar and sledgehammer to release the gudgeon pin." IOW, when that's true, apparently the nut itself is in a similar situation to the cotter pin, i.e. it's there just as a backup.

Noel

 

There is one additional factor not made clear and that is surface roughness. How was the gudgeon pin turned. Even the best lathe will produce a relativly rough finish. If I was making this component I would rough turn then grind to size. This is how taper ball joints are made for cars steering and even without laping the acuracy of the ground taper holds tight. Ask anyone who has replaced the track rod ends on a car. Now I may be wrong but I very much doubt that in the 80s preservationists had access to grinding machines with taper turning capability, or realised the need for fine surface finish. Fig 8 on p23, photo of the gudgeon pin, does not sugest the taper had a fine finish as polished high spots can be seen. I would not have enjoyed laping in such a large pin, which of course would have had to be done with they key removed then they key scraped and fitted. Polishing with emery aka James May would not have been suficient. This problem may be down to lack of the right tools back in early preservation rather than lack of inherent skills.

 

Wasn't Tangmere returned to steam in the early 2000's though via a professional restoration rather than 1980's preservation approach?

Sent from my GT-I9100 using Tapatalk

 

Hmm, I've tightened and dismantled a few large loco con rod little ends and never found one actually loose. In fact they have generally required a lot of force to both remove the nut and then to break the taper. As to the load produced by a 14 lb hammer on a 2 ft long spanner, it would be interesting to calculate the reaction of the spanner on the head of the hammer if the nut has reached refusal. For the sake of argument, assume the hammer hits the spanner at 6ft /sec and the end of the spanner deflects 0.25" Sadly my brain no longer has the ability to make the calculation but someone else may be able to. Further, the deceleration of the hammer head is greater as the spanner deflects, increasing the maximum reaction significantly. I wonder if a force of 1 ton is not unreasonable for the final part of the strike.

 

No problem in grinding the taper on a gudgeon pin. Universal Grinding machines have been around for a long time. Standard Loco drawings indicate that the gudgeon pin is ground to suit the crosshead.

 

I have re read para 40 and it gives a range of 1981 to 2004 when the gudgeon pin was made. My reference to 1980s was incomplete

 

I totally agree that the machines were generally available in the metal bashing world. My query was were they available to the preservation movement. Even today most machine tools used at steam centres are second or third hand.

 

That's a very interesting point - and it may be one the RAIB missed. (And if my guess that in normal operation the nut is a 'safety backstop' is true, it may in fact be the true originating cause of the accident.)

As discussed above, apparently the pin was normally 'locked' into place in the crosshead, when the pin was drawn into the taper under force (produced by the tightened nut). I was trying to work out how that happened, and I concluded that the taper was what did it - when loaded by inward force from the 'base' end of the tapered pin, it would have elastically expanded the tapered seat (in the crosshead) a tiny bit, and that would have wound up 'gripping' the pin (provided the taper was shallow enough - too steep and it would have tended to force it back out).

So perhaps the lack of a fine finish permitted the crosshead to (over time) lose its grip on the pin - perhaps via the embedding loss mechanism mentioned above? Interestingly, the article on embedding loss says it "can be prevented by designing mating surfaces of a joint to have high surface hardness and very smooth surface finish. Exceptionally hard and smooth surfaces will have less susceptibility".

So both parts of that may have come into play - the replacement part was made with slightly softer material (see para. 77) - although my guess is that the material issue was likely a second-order effect.


This whole thing is totally fascinating! Like the typical airplane crash these days (where usually N things have to go wrong all at once), the same may have happened here... poor finish on the pin conical surfaces caused it to come loose, then design changes on the cotter caused it to fracture...

What's equally fascinating is that I gather that the detailed understanding of embedding loss is something relatively modern - but the design of this joint appears to have been one that had been tuned to avoid the pin coming loose via this mechanism! So those 'old-timers' did in fact know a thing or two...

Noel

 

Quite frankly, it is a specialist job to grind the double taper on a gudgeon pin properly. A suitable contractor should be used.

 

Noel,

Yes, shallow tapers do self-lock when pushed in, the classic example being the Morse tapers used in machine tool spindles. They are ground to a very fine finish, and the angular tolerances are obscenely tight, although that is at least partly to down to the need for interchangeability. 1:6 is not particularly shallow though, a Morse taper is 1:20, while a CAT (or ISO, or SK, or BT) is 7:24, and is deliberately not self locking. I would say that a gudgeon pin may be difficult to remove, but should probably have axial load on it at all times. Even Morse tapers have drawbars holding them in for applications like milling, where the primary cutting load is not pushing the taper in tighter as a drill would.

Lapping two components together (the pin and the crosshead) will get you a very good finish and fit (if done properly), but the pin will almost certainly not be usable in a different crosshead. However, since 35028 seems to be running with a pin (and I would guess crosshead) from 35005 which she probably acquired about fifty years ago, the pins were probably considered to have a long enough life that this did not matter.

With the plain nut arrangement, backing the nut off against the pin is a dirty trick and should not be done, I suspect the main reason it happened in BR days was to avoid the work involved in fitting it properly while hiding the fact that the nut is actually loose. The taper does the real work, but it is always going to be preferable to have as much load as possible applied by the nut, even if you assume that the preload becomes trivial due to embedding losses with the short grip length. Think about it, if the taper does break loose you're going to have much more trouble if it can back out 0.1mm instead of 0.01mm, the radial play is 10x greater (numbers sourced from rear orifice).

I think the best arrangement for minimising the preload loss due to embedding would be a hollow pin with a separate bolt through the middle (carefully detailed to avoid fretting, naturally). Suddenly the grip length is the whole width of the crosshead, at the cost of one extra embedding interface.

Using EN3B to do EN8's job is ... interesting. I wouldn't let EN3B anywhere near that sort of job if I was designing it from scratch. However, I'd be cautious about reading too much into the state of the pin after it was violently spat out of a crosshead. The last few threads look to have stripped, probably when the nut had already worked half-way off and the pin was very loose. (Edit to add: Or they could have just had the s*** bashed out of them as it finally came out)

 

There's a lot of interesting discussion on here at the moment but no one seems to be commenting on what I consider to be the prime factor and that is the fact that the cotter was fitted parallel to the piston rod. This arrangement immediately introduces forces which the cotter is not designed nor intended to take. Yes, the cotter itself is not the best bit of design engineering but such cotters are in common use worldwide and probably without any problem. I did a quick look at quite a few little ends on locos on heritage railways this week and every one had gib-head cotters with legs split by means of a hacksaw and terminating at the point where they were opened out, usually in castellated nuts. The big difference between them and Tangmere's was that they were all vertical.

Interesting comment on the use of EN3B by 8126. The material used is not definitely known; in the report it is qualified by 'witness evidence' as being EN3B. It could easily have been EN3A, which seems to have been a favourite of BR and is specified by ORR for nuts & bolts for use in boiler work. I'm not a metallurgist but would prefer to go for something a bit more up-market; EN14A or even EN24T.