Dimensionless Space and Time

[pezman]

... Ah, the ol' Courant-Friedrichs-Lewy stability criterion, I should have
known. ...

What a coinky-dink -- the CFL coefficient was the basic problem w/ my shock-tube simulation.

[sam]

[quote]Well, I might have used the wrong speed of sound for non-dimensionalizing
the flow velocity. I used 347.2 m/sec (the speed of sound in air at 300°K).
[/quote]

Well thats probably the problem. For the true Mach Number you have to non-dimensionalise by the true speed of sound, which varies with time. It can be calculated from the temperature output file. I've attached a set of inputs files which give the correct solution, but I'm sure it'll be the same as what you already have.

Of course, the MoC is definately correct and contains no inherent errors. You can get the exact outflow Mach Number from

M = 2/(k-1) * [ (p0/p)^((k-1)/(2k)) - 1 ]

where k is the ratio of specific heats = 1.4 and p0/p = 2.83 in this case.

[WebPilot]

The very rightmost column of T.dat are the temperatures I need with which to compute
reference acoustic velocities for Xi=1.0, correct? They must be since the same number of
rows are used for time.dat and there are 10 columns of data corresponding to my 10 cells.

Here goes ...



Bingo! Good correlation between GNUDis and MoC for outflow. However, there's a 'glitch'
for inflow. The data remains shifted to the right as if no correction took place. I looked at the
T(emperature) data and there's a 'jump' here from 0.22296745E+03 to 0.29997519E+03, so
the acoustic velocity is pretty much 347.2 m/sec.

Explanation (or correction) for this?

Please don't feel 'picked on'. I have to convert my dimensionless parameters to real
values someday and I am already seeing what needs to be addressed, thanks to you.

Cheers,

-fde

PS I found a good free text editor capable of manipulating columns (blocks) of data.

[Mike Everman]

Hey Pezman, how'd you capture the excel plot?

[WebPilot]

I hadn't really given temperature variations much thought until now. NUDiS
does dump the info out in T.dat. So, I decided to plot it out in 3D and also
do a contour plot.



If you desire a zoomable color pdf file, click here.

-fde

[pezman]

I exported the data as a CSV file and imported it into the attached spreadsheet. If you look at the VBA code, you can se how the animation was done. Basically, I had a routine that called itself periodically, and it advanced the row that was displayed in the chart.

Then I used VRTainment CapturePad to grab the area of the screen that was occupied by the chart.

[Mike Everman]

Neato, Pez. That's a capture program worth $25!

And nice animation macro. I've just started macro development myself; I'll have to use that in addition to the slider. Here's a video of the NUDiS output advanced with a simple slider, as in UFLOW. Graphs are P, V, and T.

[sam]

Forrest,

[quote] I looked at the
T(emperature) data and there's a 'jump' here from 0.22296745E+03 to 0.29997519E+03, so
the acoustic velocity is pretty much 347.2 m/sec.
[/quote]

The 'jump' in temperature is physically correct and occurs when the code switches from outflow to inflow. During outflow the temperature of the gas is cold (~ -50 dec C) due to expansion, whereas during inflow ambient gas is drawn into the duct. The temperature of the ambient gas is set at 300K in the inputs.dat file.

The code cannot simulate the reingestion of exhaust products. This would require a 2D simulation in the region of the exhaust. Also, I don't believe there is much reingestion of products anyway. The literature on synthetic jets suggests that the fluid separates during outflow, rolls up into a ring vortex and convects downstream.

I don't know why your graph of Mach Number (or dimensionless velocity) is not continuous at M=0. I can only suggest that you've not done the non-dimensionalisation correctly at each point - I get a graph which is continuous. I've attached a file which contains the results of the simulation. There are 5 columns: Temperature(K), Velocity(m/s), Time(s), Sonic Velocity (m/s) and Mach Number. If you can somehow plot the final column you will see that it is continuous.


Good luck - Our results are getting closer and should converge soon I hope!,
Sam

[pezman]

I pity the poor, semi-illiterate perv who googles "Nudis animation" and ends up seeing the WMV that Mike posted.

I think that it should be possibe to put a "mask" in front of a graph -- basically a profile f a PJ with a clear interior -- and then use a graph that colorizes the data based on its value. That way, you could make animations of the density, pressure or whatever inside the jet.

My apologies to Forrest et al though for partially hijacking the thread -- I think that I'll start another if I get any ineresting results.

[sam]

Hi Pezman,
How's that shock tube simulation coming on? You can compare it directly with an exact solution, at least until the point where you start getting reflections from the boundaries. It will help to give some idea of the robustness of your code anyway.
Can I ask what numerical scheme you are using?
Cheers,
Sam

[Mike Everman]

-- and then use a graph that colorizes the data based on its value. That way, you could make animations of the density, pressure or whatever inside the jet.

I definitely want to do that. The duct profile is there, but automating the different colors will be a new learning experience!
I'll start another thread for NUDiS pre and post. Bye Forrest, you're a gracious host!

[pezman]

Sam,

I lifted the code from here:
http://www.physics.buffalo.edu/phy411-5 ... tml#topic7

I'll post the simulation as a zipped up Visual C++ 6.0 project for anyone that has the time and patience to get it working better -- it seems to have some serious issues with numerical stabilty when there are exceptionally high flow rates etc. Makes me appreciate UFLOW a bit more.

The shock tube simlation seems to work well enough, but I have had problems with simulations that have "open end" boundary conditions (e.g. the problem that Forrest originally posted).

[WebPilot]

think that this post turned out to be one of catalysis. I got a
chance to do a WD on a subject that has 'floated' on this forum for
some time now.

Larry got something out of it - 2 threads of his own branching from this
one.

Mike E. got some enhancement ideas for his pre- & post-processing of
NUDiS.

Sam got another person to give his code a try - I can get the 'amplitudes'
to match, but the 'timing' is off. Take a look at your M.jpg and see
'yours' doesn't match either. No worries, I'll figure it out someday.

And Pezman, got motivated to stitch a code together. Congratulations!

Well, it reflected and came back on me and I now wish to join the Gang
of Three who have operational codes. I got mine to begin
working, and I've dubbed it, 'ODUsF'. I was simultaneously watching the
film, "Tora! Tora! Tora!", while debugging the troublesome part.

Here is the 1st plot - it needs further testing, but it's up and running!



-fde

[WebPilot]

'ODUsF' can model a proper closed end, too. Forgive me, I am just being a
'proud' parent.



-fde

[sam]

forrest,

Care to reveal a little more about your code. What numerical scheme are you using? What boundary conditions are you including? (open and closed ends by the sounds of it). I assume its 1D - Does it allow for non-uniform cross-sections?
I've got a fair bit of experience in CFD codes and I think at times it can be one of the most frustrating things I've ever undertaken. Feeling the pain yet?
Sam