2D cup burner and FDS6

[gforney]

Please complete the following lines...

Application Version: FDS5.4.0 serial
SVN Revision Number: 4675
Compile Date: 5 Sep 2009
Operating System: Linux

The attached 2D cylindrical cup burner case runs fine with FDS5.4.0. 
However, setting FDS6=.TRUE. destroys the flow pattern and flame shape, as 
seen from the vector slices. This happens both with DNS and LES.

Original issue reported on code.google.com by jukka.vaari@vtt.fi on 2009-09-10 11:59:37

---

• _Attachment: Cup2D_CH4_Ar.fds_

[gforney]

Thanks. We'll take a look at this case.

Original issue reported on code.google.com by mcgratta on 2009-09-10 13:03:35

[gforney]

Jukka,

FDS6 has not been developed for 2D cylindrical coordinates, so this does not surprise
me.  Make FDS6=F, and then try CFL_VELOCITY_NORM=1 and FLUX_LIMITER=2 on the MISC
line.  Let's see if at least the transport scheme is working.  At one point I tried
to make that work for cylindrical coordinates.

Randy

Original issue reported on code.google.com by randy.mcdermott on 2009-09-10 14:02:53

[gforney]

(No text was entered with this change)

Original issue reported on code.google.com by randy.mcdermott on 2009-09-10 14:04:32

• Labels added: Milestone-FDS_6.0

[gforney]

The transport scheme seems to work fine. Do you mean that a future FDS6 will not be

able to handle 2D cylindrical cases?

Original issue reported on code.google.com by jukka.vaari@vtt.fi on 2009-09-10 14:19:29

[gforney]

Not at all.  Just have not gotten to it yet.  I will use this case as a test bed.

The next thing to try is to add DYNSMAG=T on MISC.  We will just keep adding options
until we figure out what is breaking it.

Original issue reported on code.google.com by randy.mcdermott on 2009-09-10 14:22:06

[gforney]

Still looking good.

Original issue reported on code.google.com by jukka.vaari@vtt.fi on 2009-09-10 14:28:15

[gforney]

Okay.  Next try BAROCLINIC=T on MISC.  Then, PROJECTION=T.

My suspicion is that PROJECTION is the problem.  Let me know.  Thanks.

Original issue reported on code.google.com by randy.mcdermott on 2009-09-10 14:30:30

[gforney]

Looks like you got it right!

Original issue reported on code.google.com by jukka.vaari@vtt.fi on 2009-09-10 14:40:36

[gforney]

Thanks for narrowing this down, Jukka.  I know where the problem is.  I'll fix it and
let you know when I have committed the new code... hopefully today.  In the mean
time, you can just run without PROJECTION, but you have to do this by setting each
option individually.  The last thing you have to set is

HRRPUV_AVERAGE = 1.E10 on REAC and you effectively have FDS6=T without PROJECTION.

Original issue reported on code.google.com by randy.mcdermott on 2009-09-10 14:47:11

[gforney]

Jukka,

I fixed the PROJECTION option to work with CYLINDRICAL (SVN 4708).  It appears to me
to be working (see attached).  Please confirm this (using FDS6=T) so that I may clear
this issue.  Thanks for catching this!

To anyone else who might be listening:

This issue brings to light the point that setting FDS6=T is not intended to give the
user an FDS 6.0 executable.  It is simply a short cut for me (and others) to test
options that would otherwise mean setting several parameters in the input file.  To
see which parameters, simply search on FDS6 within read.f90 (or search the discussion
group).  That being said, combining some parameters may not make sense.  For example,
DNS=T and FDS6=T, does not make sense, for two reasons: first, DYNSMAG=T is an LES
option and, second, for DNS I would recommend using FLUX_LIMITER=4 (CHARM) instead
of
the SUPERBEE.  The user would be better off setting the desired options individually.

One may notice that there is no accounting for a CYLINDRICAL option in the DYNSMAG
routines.  Nor will there be.  This is simply because a 2D axisymmetric calculation
is not an LES, and DYNSMAG is an LES model.  FDS will still give you an answer, but
it would be as if you simply had a 2D case in Cartesian coordinates (already dubious
enough from an LES standpoint).

Kevin,

In looking at this issue I noticed that CYLINDRICAL is used all over the place in the
code where it is not needed.  That fact that the arrays R(:)=1 and RRN(:)=1 for
CYLINDRICAL=F is sufficient.  The only place we need this flag is where we set R and
RRN and for the pressure solver.  The rest seems like we are just cluttering up the
code to save a couple of FLOPS... we have bigger problems.  Am I missing something?

R

Original issue reported on code.google.com by randy.mcdermott on 2009-09-10 20:07:03

---

• _Attachment: cup_ch4_ar_0072.png
  _

[gforney]

No, I think you're right. I was saving a few FLOPS and also just worried that 
gradients in the y (theta) direction might not be zero due to some ghost cell value

not being set properly.

Original issue reported on code.google.com by mcgratta on 2009-09-11 00:13:52

[gforney]

Randy, 

thank you for the quick fix and for the guidance regarding the various options. Nice

job. Please mark as verified.

Jukka

Original issue reported on code.google.com by jukka.vaari@vtt.fi on 2009-09-11 08:10:01

[gforney]

(No text was entered with this change)

Original issue reported on code.google.com by randy.mcdermott on 2009-09-11 12:30:26