Martin Johnson
Forum Replies Created

AuthorPosts

Hi Chris,
I begin to see your point, the nozzle may well flip sonic to subsonic and it would produce some “interesting” effects. Like you, I am eager to see just how much of an issue it is from the S160 tests.With regard to designing with DeLaval nozzles, I tend to come at the problem from a designer’s point of view (since that is what I was). Ipso, if you can avoid supersonic flow, then do so as it eliminates the issues noted above. I also note that many locomotives have minimal storage volume in the exhaust system, so there is no storage to even out the flow reaching the blast nozzle. I realise there is not much room at the front end of a loco to put such storage. However, I have been making an effort over the last few years to look at steam engine design from fundamentals and if theory dictates that there ought to be storage, then space would need to be designed in. That is also how I come to be picking holes in what others have said before – as the song goes “It ain’t necessarily so”
Martin
Hi,
I have just realised that in my second to last paragraph I give 16 inlet diameters to achieve something around 75% effy on a diffuser, that should be 8. The design chart I work from (Miller, D.S. Fig.11.3) gives the length in radii, not diameters.We are all fallible (me included as just demonstrated), However, if there are glaring contradictions in publicly quoted material, and considerable exaggeration in work by Porta it needs to be aired.
Porta gives comparisons of performance between various locomotives his Fig. 3 (His Rio Turbio 2102 being top of the league, a Geisl 9F being about mid way) – does anybody know if the curves quoted are all TESTS or is he comparing test with calculation?
Martin
Thank you both. But I am not sure you really answer the conundrum.
Chris – I am not convinced the DeLaval nozzle has an influence here. However, the time dependent regime (chuffs) is something I think is not well understood, and something I want to put a few words about on here.
John – If you look at the sidebar item “Lempor CFD?” on the page I gava shortcut to, you wil see that both CFD models are of a Lempor style arrangement, the only difference being the first example (2 plots on of pressure, one of velocity) has the blast nozzle level with the intake bellmouth, while the second (2 plots again) has it below the intake bellmouth. I cannot see how the statement by JK squares with the results the CFD plots are showing.
I think your other comments are not really relevant to the present case. But in the interest of stimulating discussion, I notice that Porta claims a typical inlet bellmouth loss of 0.04 – whereas ESDU 85032 (and my own experience of 40 years in fluid flow) suggests at least 0.05 and probably nearer 0.1.
Similarly Porta claims a nozzle discharge coefficient of 0.99 (inferred from his 1.01 factor tacked onto Equation 9). A well designed slow taper nozzle might give a Cd of 0.94 or so. What a DeLaval nozzle would give when working off design point is anybody’s guess – and DeLaval nozzles have quite a tight optimum performance band, which is hardly likely to suit heritage rail applications. Back in the day applications London to Edinburgh flat out, for example might be a different matter. What I think has happened is that Porta’s number “taken from tests at Rugby” does not account for the velocity head of steam – but I have no proof of that. I am currently struggling to reconcile EG Youngs test results which lack clarity as to where blast pipe pressure was measured and whether it includes significant velocity head.
Once again, Porta claims a diffuser efficiency of 0.8 to 0.85 with an area ratio of 4. Data in D.S. Miller – Internal Flow Systems 2 suggests you would need a diffuser at least 16 inlet diameters high to achieve 0.75 efficiency. I’d like to see that go through a tunnel!
So it is quite easy to convince everybody that your system is best if you bury some assumptions to that effect deeply enough. Am I missing something? Or does everybody else wear rose tinted spectacles?
Martin
So for a Kylchap you would have follow the above procedure for each stage, estimating values of M (entrainment ratio) for each stage taking into account that the propelling fluid is the mixture from the previous stage, and that the pressure ratio will be affected by the back pressure from the upstream stages.
Yes. Quite a challenge! But I was only trying to illustrate another way of thinking about the problem.
Maybe better to accept experimental findings that such petticoats will not beat a well designed LeMaitre or Porta setup.
Martin
I see where your thinking is going, Chris. My own thinking was along the lines of changing the capacity of the smokebox, the resistance of the tube bank and the capacity of the firebox. I think in your electrical terms it would be a variable voltage feeding a capacitor connected to another capacitor via a resistor. Either way it is a differential equation of second or third order. I think that the frequencies involved are way below any possible resonance, so it would reduce to a mass oscillation problem.
Oh how I wish my degree level maths was still up to standard!
Martin
Did those results ever get processed and posted somewhere? A variation of nearly 2:1 in vacuum, presumably in time with the “chuff”, sounds interesting. What implications might this have for design guides of smokebox volume, tube bank flow resistance and susceptibility to fire lifting / clinkering?
Martin
Thanks for that link, Chris. You have partly saved me the trouble of replying. The Thermopedia data follows very closely the approach in the Engineering Science Data Unit Report 85032 – Ejectors and Jet Pumps Design and Performance for Incompressible Flow. There is another report covering compressible flow (sonic choking of blast nozzle) where the maths gets even heavier – Oh joy of joys!
It is report 85032 on which I based my comments. Incidentally, as far as I can untangle the maths, the basic approach is the same as Porta uses.
ESDU 85032 contains the following graph:
ESDU Report Fig 3awhich assumes certain typical loss coefficients for the nozzle, secondary inlet and diffuser, and also assumes that density of the motive fluid (exhaust steam) is equal to the density of the entrained fluid (combustion gas) which is approximately true. It shows the envelope of best basic design against volume ratio and pressure ratio.
The graph is expressed in terms of volume ratio (gas / steam) and pressure ratio (draught / exhaust back pressure) and area ratio (blast pipe / choke). For a typical locomotive M is around 2 (varies with degree of superheat, grate area, working point etc. etc.)
You will see that for M = 2, travel across to the dropping “M” line and you are well to the left of peak effy. You will also see that dropping down to the x axis suggests an area ratio of 0.12. With that set up you can expect an N value around 0.15. You would then need to size a blast pipe to give a back pressure 1/0.15 times the total draught. The choke would be 1/0.12 times the blast pipe area.
The calculation would then be repeated using refined values for individual losses and density ratio etc. to optimise the design.
I have recently waded through the Everett G Young paper which seems to confirm much of what a retired fluid dynamicist knows about diffusers, settling lengths and maximum diffusion angles. There are some very subtle design compromises to be made in terms of choke geometry and diffusion geometry to get the best out of a chimney, much of which will be down to experimental results – not calculation.
Martin
Thanks for that link, Chris. You have partly saved me the trouble of replying. The Thermopedia data follows very closely the approach in the Engineering Science Data Unit Report 85032 – Ejectors and Jet Pumps Design and Performance for Incompressible Flow. There is another report covering compressible flow (sonic choking of blast nozzle) where the maths gets even heavier – Oh joy of joys!
It is report 85032 on which I based my comments. Incidentally, as far as I can untangle the maths, the basic approach is the same as Porta uses.
ESDU 85032 contains the following graph:
[url=https://flic.kr/p/27kvbE5][img]https://farm1.staticflickr.com/912/42227540672_a30511930c_b.jpg[/img][/url][url=https://flic.kr/p/27kvbE5]Fig 3a ESDU85032 copy[/url] by [url=https://www.flickr.com/photos/140734312@N06/]Stan Wellbach[/url], on Flickr
which assumes certain typical loss coefficients for the nozzle, secondary inlet and diffuser, and also assumes that density of the motive fluid (exhaust steam) is equal to the density of the entrained fluid (combustion gas) which is approximately true. It shows the envelope of best basic design against volume ratio and pressure ratio.
The graph is expressed in terms of volume ratio (gas / steam) and pressure ratio (draught / exhaust back pressure) and area ratio (blast pipe / choke). For a typical locomotive M is around 2 (varies with degree of superheat, grate area, working point etc. etc.)
You will see that for M = 2, travel across to the dropping “M” line and you are well to the left of peak effy. You will also see that dropping down to the x axis suggests an area ratio of 0.12. With that set up you can expect an N value around 0.15. You would then need to size a blast pipe to give a back pressure 1/0.15 times the total draught. The choke would be 1/0.12 times the blast pipe area.
The calculation would then be repeated using refined values for individual losses and density ratio etc. to optimise the design.
I have recently waded through the Everett G Young paper which seems to confirm much of what a retired fluid dynamicist knows about diffusers, settling lengths and maximum diffusion angles. There are some very subtle design compromises to be made in terms of choke geometry and diffusion geometry to get the best out of a chimney, much of which will be down to experimental results – not calculation.
Martin
Not sure if the above link will work on here, so try this one if there are problems:Fig 3a ESDU 85032
I had come to the conclusion that the combined input area of the various bellmouths in a Kylchap system would be greater than for any other common exhaust system
Yes, they have to be Petticoat 2 takes input from Petticoat 1 & some flue gas. Petticoat 3 takes input from Petticoats 1 & 2 plus some flue gas.
Martin
Chris,
My comment about efficiency was more in relation to where the peak efficiency of a jet pump tends to lay relative to where it is used in a locomotive. As a general rule, for the mass flow ratio (flue gas/steam) found in a coal fired loco you are well to the left of the effy. peak. One way to get round that is to take the flue gas in stages. So the first petticoat is working at a more favourable mass flow ratio (hence greater efficiency), as is the second etc.The net result should be more draught flow for less steam back pressure – always the goal. As an added bonus you get a better flow distribution across the tubeplate!
I haven’t worked the numbers, but surely sonic limits would only be an issue at the blast pipe orifice – everything else is working at much lower velocities. Another point I only realised this morning is that the flue gas is cooling as it flows through a chimney system, due to mixing with the exhaust steam. Therefore, the flow is decelerating even in a paralell pipe, because the density is increasing and hence volume is decreasing for a given mass flow.
Perhaps another way to look at things, which addresses some of the other points you raise, is that surfaces usually help to control diffusion of flow and uncontrolled diffusion is the enemy of efficiency. I am afraid that enigmatic statement has taken a career in fluid dynamics to really sink in.
Martin
Made a rather hurried reply to Chris’s query yesterday. Should also have said that if each petticoat has significant contraction, then you will need more pressure energy to drive the consequent increase in velocity energy – which is also not a good thing for the stage preceeding. Any interchange of velocity / pressure energy is lossy (pressure to velocity slightly so), hence leave the petticoats parallel.
Martin
Chris,
The way I come to that problem is that the mass flow ratio in a locomotive is not in a very good place for efficiency in a jet pump. Obviously, we can’t change the mass flow ratio very much, coal combustion does what it does.That being the case, one way of overcoming that is to stage the pump, so we have a blast pipe that entrains a small proportion of the flue gas flow, this then feeds to another stage where the flow entrains another load of flue gas, and so on. Each stage is thus working at a more favourable mass flow ratio.
If the flow in each Kylchap petticoat is being used to drive another stage, then we do not want to decelerate the flow, as we need that velocity energy to drive the following stage. Hence we could either use a slightly converging or parallel petticoat.
Martin
P.S. Glad to have stirred things up a bit!Well yes but………….
I am not sure if there are any other explanations, but I could not get any sound on the link, but it looks to me as though the Reynolds number is very low in this visualisation – I think we are looking at the Laminar / Turbulent transition as the flow regime changes shortly after the nozzle to the vortex producing turbulent. At low Reynolds numbers, viscous forces dominate and it is viscosity that generates vortices – vortices are impossible in a completely inviscid fluid.Real blastpipes will be working at high Reynolds numbers, not the lazy flow you see in the video.
Also note that the accepted theory of blast pipe design relies on Bernoulli, Momentum exchange and Conservation of Energy – notice that list does not mention vortices.
The case of the “Jimmy” is also misleading, because the Jimmy is probably reducing efficiency of the system, but such devices also increase back pressure and hence the energy input to the system and hence the overall performance increases. That is not to be confused with well designed multiple nozzles, which will have no or very marginal efficiency drop.
So while the flow visualisation is interesting, I would be wary of reading too much into it.
Martin
I notice the Indian abstract refers to
a recirculation zone near the primary nozzle exit
That seems to imply major flow separation, which can only happen with uncontrolled diffusion – an inherently unstable process. When I stopped real work, the state of the CFD art had not cracked uncontrolled diffusion. Has it moved on in the last 10 years or so?
Pulsations – 3 ways round that:
a storage volume in the steam exhaust – not much space “up front”
larger blast pipe orifice – may not be feasible to achieve smokebox vaccum.
Graduated exhaust opening on the valve system – may impose too much exhaust back pressure.
Take your pick.Martin

AuthorPosts