Boiler design & construction ex Patriot thread.

I think the new build stuff I recall runs at about 150 or 180 psi. I've a feeling the steam pressure/temperature function of a saturated steam vessel is a fixed curve (i.e. a given pressure will always equal a given temperature).

As for dodgy boilers - I'm not going to start naming and shaming, but some boiler inspectors (one who has recently retired was popularly known as "blind ____") and some of the organisations operating some of the more tired stuff out there certainly give me some cause for concern.

In my opinion, part of the problem relates to the fact that some insurance companies insist on using their own inspectors, who aren't always overly familiar with railway type fire tube boilers. Some insurance company inspectors are truly excellent, know exactly what to look for and where to look, however others spend most of their year looking at coffee machines and garage air compressors, and when presented with a riveted constructed fire tube boiler with a copper firebox are clearly out of their depth.
It cuts both ways - some inspectors are (IMHO) rather over zealous - I've certainly replaced stuff in the past I'm fairly sure would have done many years more before it croaked!

My view of all this is probably slightly jaundiced - I work as a boiler-maker, therefore most of the stuff the I work on has been run till there is no more running to be had from it - that's why I've taken it to bits.

 

Correct, Boyles law and steam tables. http://www-mdp.eng.cam.ac.uk/web/library/enginfo/cueddatabooks/thermofluids.pdf

Is there some sort of handbook or reference guide the inspectors can work from to guide them a bit more?

 

"Finite element analysis may well allow better understanding and design of boilers. Does your analysis include for the effects of temperature and expansion or is it purely looking at stress and strain?"

I did do a rough and ready temperature analysis based on a 100 degrees C delta of the firebox when compared to the wrapper. The plots show the stresses due to the temperature delta on the left, the pressure in the middle and their combination on the right. The largest differential expansion stresses are concentrated in the end rows of stays and foundation ring corners.

The boiler is an approximation to the all welded, all steel LNER 100A design from Alan Haigh's Locomotive Boilers book. The results are in N/mm2 and the choice of colour in the plots is somewhat arbitrary. Roughly speaking, blue/green is within design limits for the design standards, yellow is above those limits but below material yield and red is getting to the point of plastic distortion. The FEA program used does not however model plastic deformation which may relieve stress levels in real life.

 

It is possible to look at the loads on individual stays in the model shown above. I chose a couple that I thought might be most affected by differential expansion and got some counter intuitive results below. What is intuitive is that making the stays flexible at the wrapper significantly reduces bending stresses. The bending stresses dominate when compared to the tensile load that the design standards assume. The fact that the largest stresses are in compression needs a bit of thinking about .

I won't post anything else for now but would welcome comments questions if there is any interest in further investigation.

 

It's certainly interesting, but I have to admit I'm not totally convinced by your results - I'm pretty sure stays are actually in tension pretty much all the time (if they weren't, why do fireboxes bulge inwards when the stays fail in an area?) - I suspect that there is something your model is missing, or some flaw in your parameters that accounts for this.

One would expect differential expansion to put some compressive load onto the stays, but the effects of the boiler pressure should more than exceed this.

That stays have issues with sideways loading has long been known - one of the reasons oversize stays are usually "waisted" (turned down to the original core diameter) between the plates is to ensure they are at least as flexible sideways as the originals.

 

I have to admit to being suspicious of the stay that appears to be in compression as well.

I've plotted consecutive alternate central crown stays (welded at both ends) in the diagram below starting with CCS1 (Central Crown Stay 1) and showing every alternative stay up to CCS15 (The x axis of each graph starts at the wrapper)

The full boiler model is a shell and beam model and takes a bit of getting used to. A fully detailed solid model of the whole boiler would take weeks to solve on my ageing laptop and I am limited to producing bread slice solid sections as shown on the right hand side of the diagram. In the full boiler model each stay is represented by a line. These appear in the zoomed in section of the firebox crown below but they are quite difficult to see and the plots can only provide limited information ie the maximum stress.

I've exaggerated the distorted shape prediction shown below the graphs so that it shows a 25 fold exaggeration of how the boiler is deforming when under pressure and expansion loads. I've added the plot of the full boiler in a separate post below this one.

Looking at this I can begin to believe that tubeplate expansion is putting the front row of crown stays into compression. The bending is greatest at the firebox end because the crown is bending downwards but these stays are 2 ft long and comparatively flexible. The wrapper is not that distorted so the bending load at the wrapper end of the stay is less and the compression caused by tubeplate expansion dominates. The other stays are as one would expect and fairly consistent away from the ends of the firebox.

Does this make any sense?

 

This is the full boiler model showing distortion exaggerated 25 fold. The picture shows a half section viewed from inside.

 

It's normal practice on boilers of any size (I.e. most standard gauge industrials up) for the front two or so rows of crown stays not to be direct stays, but some sort of flexible link stay. My understanding is that this is to stop the crown stays failing prematurely due to expansion stress.

That said, I've never opened up a boiler where these have been anything other than seized solid.

Are you able to model anything like that?

 

Something like this? Definitely worth doing and I will have a go but might take a few days to get a sensible model. Looking at the arrangement the stays appear to go through a conical hole in the upper support and the upper bolt can lift a little from it's seat. It's quite tricky to model some of the things that go on in a real boiler in a computer approximation - computers are no substitute for experience and real metal. It is however cheaper to build a computer simulation of a boiler than it is to build the boiler itself!

 

That's the sort of thing - the holes in the links are usually parallel clearance holes, rather than anything more fancy. The other arrangement you sometimes see has flat plate links secured to the top of the box with a second pin parallel to the one in the crown.

What package are you using? I use Autodesk Inventor a lot at work, but not really played with the FEA bits of it...

 

It's an old FEA package called Mechanica. The company that wrote it was taken over by PTC. I have a pretty old version from when I used to work for the original company which is functional but very unforgiving. It is not unusual to work for a hours only to end up with a problematic model and have to start again. It does work very well when it works though and was initially written as an Engineer's tool to be used as early as possible in any new design - when it's cheap to change things.

I've used Inventor to help with the design of the new boiler for the fast umpire launch I help look after. The 3D model transfers really well to Onshape making it easy for others in the group to suggest improvements. I've not used Inventor's FEA software - meshing the fully detailed solid model would make a very unwieldy FEA model.

I'm hoping we can laser cut all the plates direct from CAD data. We've just had approval for the design itself.

 

Be careful doing that. While laser cutting is very good, it tends to put a hard skin on the edge of the cut line. A few years back we built a brand new traction engine boiler for a customer, and had the hornplates laser cut, including all the stay holes, which were cut slightly undersized, so we could ream them truly round.

What we found was that the edge hardening from the laser cutting made reaming the stay holes out almost impossible (we were using a bridge reamer in a big air drill), the reamer just didn't want to cut the edge.

If you are looking at making anything similar, I would suggest investigating water jet cutting instead, which doesn't suffer from this problem.


Having had a very quick look at your launch boiler design, I must confess I'm less than keen on square cornered firebox shells and throat plates, and I fear your design hasn't got enough water space between the inner and outer fireboxes for a square cornered design.

Having opened up a few square cornered boilers like this which have been in service for a few years, I've seen evidence that they often suffer from poor water circulation round the lower half of the firebox, as the sharp corners restrict the water flow, particularly the flow up the throat plate and into the barrel.

This can cause the water round the firebox to get so hot it starts to break down into little bubbles of hydrogen and oxygen. Some of the hydrogen gets driven into the plates and makes them become brittle, while some of the oxygen works as an accelerated rusting agent, and thins the plates down. Eventually the thining and embrittlement result in the plates cracking.

 

Never heard this theory before, John. It also contradicts an earlier (and correct) statement in post 41 where you said: "I've a feeling the steam pressure/temperature function of a saturated steam vessel is a fixed curve (i.e. a given pressure will always equal a given temperature)." Water can't get hotter than its saturation temperature. Above that, it is steam and, further above that it is superheated steam.

 

I phrased this badly ^^^^ - what seems to happen is that the flow round the box can go wrong to the extent that you get localized pockets of steam (indeed super-heated steam) effectively trapped against the firebox plates (the water flow cannot carry the heat away fast enough, so you get a localized layer of steam between the firebox and the water), some of which then break down at a molecular level, and causes the hydrogen embrittlement, and accelerated corrosion.

I'm not going to name and shame on here (ask me the story next time you are down at our place), but a few years ago, we were involved with an official investigation of a near miss with a square cornered boiler where we found a lot of these shortcomings, and concluded that all the evidence pointed to the design having waterways with insufficient space and sharp corners which had caused this effect ^^^ - the firebox sides had severe cracks, and we got some of the plate analysed, and the metallurgists said hydrogen embrittlement was the cause.
We redesigned the boiler with a sloping throat-plate with a pressed throat and corners, to encourage more flow from lower down the firebox, and from the evidence to date, it appears to have worked.

 

Only just caught up with this thread.

Too late, you just have! Lol
Sorry to be the bearer of bad news, but the individual concerned has not retired - although that was a widely circulated rumour last summer/autumn time when many people struggled to contact them.
They are still active today (or at least were as of 2 months ago).

I have my own views on their attention to detail - as a result of 'inheriting' them 10 years ago as the existing inspector for my boiler.
(....I swapped inspectors before their ticket expired.)

 

This is a first attempt at swing stays. It's taken a few hours to get the model to work after modifying it and the results are partly nonsensical at this stage. Component thickness doesn't show in this type of FEA model. The horizontal plates that the pairs of stays are suspended from are 25mm thick while the suspension web is 12mm. I'll try tuning both of these to see how the results change. each stay is free to rotate in the top support but welded to the firebox crown.



The two graphs show the maximum stress value along the stay length for the original central(ish) front fixed stay and 1st attempt swing stay. The swing stay carries much lower stress but is still in compression because of tubeplate expansion. At the moment the bracket is forcing the second row of stays down and that doesn't strike me as good. I suspect there is more work to do to produce something sensible.

 

After a few frustrating software crashes (20 year old software) I've managed to put together some different stay support variants for the front of the firebox and still can't get anything other than compression in the first row of stays.

So some simple sums follow which I hope I haven't goofed?

If we assume the firebox tubeplate is 1 m tall and at 300 C, it will expand by about 3 mm. Further assume that the front vertical stay is 0.5m tall and at 200C it will expand by 1 mm. The top end of the stay will be pretty determined to move upwards by a total of 4 mm.

If we assume an idealised vertical wrapper ( adjacent to the stay with no opportunity for beam bending) is 1.5m tall and at 200C it will expand by about 3 mm. If we further assume it is loaded to 100 N/mm2 it will stretch a further 0.5 mm. A total of 3.5 mm. So the front vertical crown stay will be pushing up against it.

The tubeplate will move forward as well as up so a more complicated calculation might still introduce some tension but the idea of stays near the tubeplate being in compression might not be as unlikely as it would first seem. Also the crown stays in this particular theoretical boiler design do allow the top of the wrapper to be pulled down about 4mm in the middle by the firebox crown. This can be seen in the pictures above

The picture below (I should have kept it as a thumbnail but it took so long to produce it deserves to be shown in full) shows four different support scenarios for the front two rows of crown stays. Given the fact that the front row are in compression the only support they give is via bending and the sling arrangement only serves to reduce that bending.

Leaving the entire front row out altogether does not disastrously affect the boiler and opens up the potential for sensitivity studies around exactly where to position the front row of stays for best effect. It would also be interesting to look in more detail at a curved corner firebox and see how the position of the stays affects the stresses. It would also be worth looking to see how locally reshaping the wrapper could help compensate for the expansion effects. For example it the wrapper was flattened slightly over the firebox tubeplate the pressure forces would tend to lift it again to form a circle and help maintain support for the front of the firebox.

So many things to try so little time...

 

Does anybody have drawings of a Diagram 118 boiler? The model above is based loosely on a welded variant LNER Type 100A boiler but I haven't got the throat area right. I'd like to build a new analysis model with a wide firebox and rounded wrapper and firebox corners.

 

Try asking on the P2 or Tornado threads for a PM... ?

 

Thanks,

The analysis system is not up to creating the required compound curves so I'm trying a two stage process, modelling the plates and superheat tubes in a modern 3D CAD system and then completing the FEA model by adding all the stays and small tubes in the analysis system. This is a bit of a grand challenge as the resulting model is going to be big, take a long time to build and there is no guarantee that it will work.

I managed to find an old LNER 2-8-2 GA drawing and derived a generic wide boiler design from it as per the first picture below. Hopefully this looks fairly sensible? The original 3D CAD model is viewable here . The link takes you into a very nice (ie free and capable) 3D modelling system called Onshape. Holding down the right hand mouse button whilst moving the mouse over the CAD model in Onshape allows you to move the camera view around.

The model should be at mid plate dimensions because the analysis system treats each plate as a surface with a thickness attribute. It's what is needed for the analysis model rather than a fully detailed production model.


The CAD model reads into the FEA system ok as per the second picture below.