Nice Curves for a Short Lady?

[larry cottrill]

Bill H, Mike E, Graham (others welcome, too, of course) -

What do you think of the following UFLOW1D pressure curves for a simple closed pipe engine 26 inches long?

L Cottrill

[larry cottrill]

When I ran this, I couldn't believe it -- I thought I must have accidentally changed some parameter that I didn't want to modify. So, I went back by removing this change, and got exactly the same pressure curves as before.

What it amounts to is this: This engine design responds incredibly well to the addition of a modest tailpipe flare. The earlier curves are from a 1.25-inch ID tailpipe, with no tail flare. Here are the curves when a simple flare to 1.5 inches max diameter is applied. The overall length is exactly the same, and no other changes were done. The flare takes up about 1/3 of an inch of the length of the pipe.

I find this result amazing ...

I went back and tried applying this to the Elektra I. There was nothing like this effect -- the only observable difference was that the Elektra I pressure trough was depressed slightly farther below the ambient pressure line. The shape of the curve was essentially unchanged. So, the effect shown here may be unique to this type of design.

It does suggest, though, that a small flare can have a large impact on the breathing of an engine, at least in some designs. It is obviously well worth trying at some point in the development cycle.

L Cottrill

[jmhdx]

Not sure what we're seing there Larry but isn't that just the thing with Uflow, unpredictable and unexplainable results can be obtained and I'm sure evidence supporting any theory you'd care to mention could be generated with it. I trust it with simple pipes and simple pressure changes but don't feel its up to the job of pulse combustion events. There are others who have perciviered and maybe some conclusions will be possible. I have concluded that the tailpipe needs to be conical, but thats all.
Thankfully your a builder and a talker and I hope you can soon evaluate the usefullness of Uflow in predicting actual events.
Mike.

[larry cottrill]

Not sure what we're seing there Larry but isn't that just the thing with Uflow, unpredictable and unexplainable results can be obtained and I'm sure evidence supporting any theory you'd care to mention could be generated with it. I trust it with simple pipes and simple pressure changes but don't feel its up to the job of pulse combustion events. There are others who have perciviered and maybe some conclusions will be possible. I have concluded that the tailpipe needs to be conical, but thats all.
Thankfully your a builder and a talker and I hope you can soon evaluate the usefullness of Uflow in predicting actual events.
Mike.

Mike -

Of course, using the program has shown that slightly tweaking some parameters skews results considerably. However, we need to be careful not to assume that apparently odd results are simply wrong.

Keep in mind that what I did here was not a change to any temperature parameters, either of the pipe or the gas. Those seem to be the ones that yield thrashing results if not reasonably managed. What I did here was make a minor alteration to the pipe geometry -- simply curling out a small flare right at the end of the pipe. When I try that with Elektra I, the difference in the pressure wave at L/8 is barely noticeable. With the 26-inch Elektra II, however, the effect is larger: the low-pressure trough is 'magnified' to a visibly lower curve that should create better breathing, all other things remaining unchanged. See my latest post to the 'Elektra II ... Revised ...' thread to see the effect. I think it seems quite reasonable that a small tweak in the geometry should show up like this.

Here, I've done the very same thing. Nothing has changed except curling out the end of the tailpipe a few mm . The overall length is the same as in the E II -- but look at the difference in the overall effect! Unless someone can show me a flaw in my approach, I have to conclude that there is some reason that the geometry of this engine is simply far more susceptible to such a seemingly minor tuning 'tweak' than the cruder Elektra design.

The Elektra series is designed primarily for ease of construction; this one is not, but is designed according to my latest hypotheses for gaining engine performance. I think this results in a finer tuned mill that senses the small things to a lot higher degree. Now, I'm sure that also means that small flaws in construction would have a much higher impact on performance as well!

Of course, I could just be wrong ...

L Cottrill

[sam]

Larry,
I'm afraid I agree with Mike - I'm almost certain you're seeing a "numerical artifact" as people tend to call them. You have to remember that UFLOW is essentially a 1D flow solver and fiddled to allow area changes. I've done a fair bit of research using 1D simulations and they tend to break down when there are sudden geometry discontinuities; an intake flare is a good example of this. I would suggest increasing the number of computational cells in the domain quite considerably and see if you get the same result. I'd be very surprised if you do.
sam

[larry cottrill]

Larry,
I'm afraid I agree with Mike - I'm almost certain you're seeing a "numerical artifact" as people tend to call them. You have to remember that UFLOW is essentially a 1D flow solver and fiddled to allow area changes. I've done a fair bit of research using 1D simulations and they tend to break down when there are sudden geometry discontinuities; an intake flare is a good example of this. I would suggest increasing the number of computational cells in the domain quite considerably and see if you get the same result. I'd be very surprised if you do.
sam

Sam -

Well, of course, you guys are probably right. However:

- I am using a 'mesh size' of 2 mm -- far smaller than the length of the flare I added.
- Recall that there was no such "wild" effect in working with the Elektra forms of the same or nearly the same length. The same mesh size was used in those calculations, too. Also the same tube temps, basically, with just a little difference in how the temps are distributed, due to the variation in engine geometry.

L Cottrill

[larry cottrill]

Larry,
... I would suggest increasing the number of computational cells in the domain quite considerably and see if you get the same result. I'd be very surprised if you do.
sam

OK, I re-ran it with mesh sizes of 4mm [fairly quick] and 1mm [crawls] -- the resulting curves are essentially identical. I don't think I can go bigger than 4mm because one piece is only 4mm long in the pipe table.

L Cottrill

[sam]

As promised, I'm very surprised.
I still think its essentially wrong but can't think exactly why at the moment. If you tell me what the exact geometry of the pipe is and your initial conditions and boundary conditions then I'll run the same simulation on another piece of software I have and see if they compare. send me a private email to sam_mas0n@hotmail.com if you want.
sam

[larry cottrill]

Well, it turns out that it can be run successfully with an 8mm mesh, giving the same curves in almost no time at all.

You might be interested in how UFLOW1D sees this engine operating, and then re-evaluate whether you think a pipe can exist that actually works something like this in real life. So, here goes ...

In the ...at_start graph, we see the starting condition. I have limited the fully pressurized area to about the front 20 or 25% of the chamber, right up against the front wall. The steps in the density graph are due to the pipe and gas temperatures I have put in for the different sections of the pipe. I think it is reasonable that the pipe runs cooler the farther rearward you get. I am also assuming that there is some cooling as the intake flow piles up against the front plate before explosion.

The ...at_1 graph is approx. one ten-thousandth second later. The pressure wave is well-formed and moving away from the front wall. There is the appearance of reasonable velocity and mass flow. This pressure wave shape is modified quickly as the wave moves through the chamber toward the straight pipe. The tiny anomaly at the very left end is due to "spark plug leakage", i.e. the fact that you have to have an open end for UFLOW1D to work. This is a 1mm diameter hole; anything smaller breaks the run.

The ...at_2 graph shows the pressure wave once it gets into the tailpipe section. A perfect 'sharkfin' wave with an extremely fast-rising leading edge and almost flat ramp for a trailing edge. As we would expect, the wave will hold this shape for almost its entire traverse of the pipe. The gas velocity and mass flow are now well developed. Time is 0.0004 second.

In the ...at_3 graph, the picture is much the same, except that a standing wave has been fully developed in the chamber zone. Note the extremely small standing wave magnitude achieved because of the careful nozzling of the chamber. Time is now 0.0006 sec.

In the ...at_4 shot, at about 0.0008 second, the pressure wave is about to encounter the small tailpipe flare [from 35mm ID to 38mm ID in 8mm length]. Note that the pressure, velocity and mass flow curves are essentially the same as before.

But, wait till you see what happens next!

[to be continued, next post]

L Cottrill

[larry cottrill]

[continued from my last post]

OK, now the fun begins!

In the ...at_5 shot, everything changes drastically. At this point [approx. 0.0009 seconds, only 1/10,000 of a second later than the last graph], the gas in the tailpipe flare snaps suddenly into supersonic flow! The pressure in that zone plummets, pulling most of the pressure wave down with it. There is, naturally, a sudden drop in gas density and an extreme jump in mass flow in the region at and forward of the flare. A true deLaval nozzle flow. Yes, unbelievable!

But wait ... there's more: 2/10,000 of a second later [the ...at_6 graph] we see extremely rapid formation of the pressure trough, and the beginning of its advance forward through the pipe. Mach velocity and high outward mass flow continue in the region adjoining the flared end of the pipe. The pressure trough will develop steadily through the pipe from tail to front end from this point on.

The ...at_7 graph is much later, at 0.0021 second. The pressure trough is a ramp of decreasing pressure that continues to grow from tail to nose. Most of the velocity in the pipe is now sub-Mach, appearing as an astonishingly smooth distribution from about zero at the front end to just under Mach speed where the flare begins, then jagging up rapidly to almost the same supersonic velocity as before at the pipe end. The velocity is mirrored in the mass flow graph, an even smoother distribution. At this point, the density in the whole pipe is very low and almost uniformly distributed throughout. The fairly low values of mass flow are a result of this extremely low density. Yet, the flow is still outward, as if the high velocity in the flare is acting as a blocking valve!

At 0.0038 second, the ...at_8 graph shows the entire pipe uniformly depressurized to a low value! The density throughout the pipe continues to drop, with the massflow all still outward at a very low set of values, while gas velocity is almost the same as before! Again, the velocity in the flare remains high, seemingly blocking encroachment of ambient air from the tail!

Finally [you thought it would never end, didn't you?], at 0.005 second as shown in the ...at_9 picture, the pipe is almost fully purged of gas mass from the explosion. Pressure, density and mass flow are very low, with mass still exiting the pipe at high velocity! No sign of any air entrainment from the tail end, even now.

Oh, my friends, if the world could only really work like this!

L Cottrill

[jmhdx]

Well I'm very impressed with your work and I believe youv'e provided sufficient evidence,at least as far as Uflow is concerned, to support your findings. This would appear a major development and hope you can put your name to the effect before someone else claims to have observed this previously."Cottril check flair"?
So what can we summise?, that the pressure differential created by the flair allows a build up of the remaining mass with sufficient velocity to resist the atmospheric pressure beyond the pipe for much longer allowing a greater vacumn to develop? Please offer a one scentence theorem if you can.
As we have not seen the inflow stage of this cycle I have to ask if you do have the pressure at the pipe end set to 100000pa or one atmosphere?
Regardless of this I hope your busy sticking a flair on the back of your running Elektra.
Mike.

[larry cottrill]

Well I'm very impressed with your work and I believe youv'e provided sufficient evidence,at least as far as Uflow is concerned, to support your findings. This would appear a major development and hope you can put your name to the effect before someone else claims to have observed this previously."Cottril check flair"?
So what can we summise?, that the pressure differential created by the flair allows a build up of the remaining mass with sufficient velocity to resist the atmospheric pressure beyond the pipe for much longer allowing a greater vacumn to develop? Please offer a one scentence theorem if you can.
As we have not seen the inflow stage of this cycle I have to ask if you do have the pressure at the pipe end set to 100000pa or one atmosphere?
Regardless of this I hope your busy sticking a flair on the back of your running Elektra.
Mike.

Mike -

Yes, the ambient air is at the standard pressure, as usual.

Unfortunately, UFLOW1D doesn't show this kind of effect with either the Elektra I or II -- all you get is a somewhat improved suction wave, but the improvement is minor [though welcome, of course]. It takes the right shape to drive the gas mass sufficiently to produce the kind of radical effect shown here.

Of course, the reality may come a lot closer to what the program says happens in the Elektra engines than what we observe here. Maybe we've crossed some sort of threshold where the numbers 'thrash' in a way that just happens to look good to me in this particular case.

Of course, that doesn't mean it wouldn't be a beast worth building, even if UFLOW1D turns out not to describe the reality.

If it does work like this, though, you really have to re-think what a 'pulsejet' is, or else call this something else. Here's why: This engine apparently never re-builds the tailpipe 'piston mass' from cold ambient air; the only breathing it wants to do is through the intake pipe [unless breathing there puts the brakes on the process so fully that it can completely kill the action and allow the low pressure to pull air in]. If, in fact, the piston mass is never replaced, only the power of the pressure wave is available for thrust. In that case, we have essentially a 'wave engine'. Such an engine will have to produce a VERY powerful pressure wave and [probably] cycle very rapidly to generate reasonable thrust. Since we apparently get decent pressure trough development by about .003 second, it seems to me that we could hope for about 300 hz operating frequency, at best.

L Cottrill

[larry cottrill]

As promised, I'm very surprised.
I still think its essentially wrong but can't think exactly why at the moment. If you tell me what the exact geometry of the pipe is and your initial conditions and boundary conditions then I'll run the same simulation on another piece of software I have and see if they compare. send me a private email to sam_mas0n@hotmail.com if you want.
sam

Sam -

Since you have responded, I would like that. I am interested in the development of this, if it does anything like what I show in the graphs I posted. Can you accept emails with jpeg attachments? Do you know what the limit is on total attachment size in a single email?

[NOTE: It is not my intent to keep this geometry "under my hat" for too much longer, so eventually the whole forum will benefit. I just thought that even a short lady deserves to have an air of mystery about her.]

L Cottrill

[Mike Everman]

Please offer a one scentence theorem if you can.

Mike.

Mike, you do realize you're talking to Larry, right? Ha Ha ;-P

[Nick]

Can i just say on a fairly irrelevant note that I'm pleased to see Sam and JMHDX aboard as UK contributors, sometimes I feel Graham and I are the only mad (enlightened?) people on our little island.
Just got to run my head under tap for a bit now to cool the grey matter after reading the previous posts. :-)

Nick