Numerical Instability due to sudden HRR increase

[pxlbarrel]

I'm getting the "numerical instability" error and it looks like it's
being caused by a sudden increase in the heat release rate (about 500
kW to 3 MW in one time step). Up until the error, the results were
promising (ie, doing what I expected it to do). It is a vitiated fire
(the FDS HRR curve is slightly below the calculated HRR). I planned
it as a vitiated fire.

The main purpose of the exercise is to determine at the point when
oxygen decreases to a point when combustion can no longer be
sustained. My expectation was that it would have a jump in heat
release rate and then start dying down almost immediately because of
the depletion of the oxygen. However, it would seem that the jump
that I was expecting is causing it to stop, probably precious moments
from when I would have stopped the simulation myself.

I've got a couple of pieces of upholstered furniture ignited with a
burner that I turn off after a little while (flame continues to
propogate after the removal of the burner). There is no external
source of oxygen. Only one mesh. Using the most current version of
FDS. Fails in both XP and Vista (desparation dictated the running on
two platforms).

Here are the last 3 entries in the .out file:

Time Step 18600 September 21, 2009 17:15:10
----------------------------------------------
CPU/step: 1.800 s, Total CPU: 8.95 hr
Time step: 0.01542 s, Total time: 410.96 s
Max CFL number: 0.92E+00 at ( 14, 57, 14)
Max divergence: 0.39E+01 at ( 21, 58, 6)
Min divergence: -0.62E+01 at ( 16, 58, 14)
No. of Lagrangian Particles: 499700
Total Heat Release Rate: 557.647 kW
Radiation Loss to Boundaries: 319.251 kW

Time Step 18700 September 21, 2009 17:18:11
----------------------------------------------
CPU/step: 1.803 s, Total CPU: 9.00 hr
Time step: 0.01477 s, Total time: 412.47 s
Max CFL number: 0.89E+00 at ( 14, 57, 15)
Max divergence: 0.35E+01 at ( 26, 56, 5)
Min divergence: -0.44E+01 at ( 15, 58, 5)
No. of Lagrangian Particles: 499699
Total Heat Release Rate: 557.935 kW
Radiation Loss to Boundaries: 314.080 kW

Time Step 18749 September 21, 2009 17:19:42
----------------------------------------------
CPU/step: 1.858 s, Total CPU: 9.02 hr
Time step: 0.00001 s, Total time: 413.23 s
Max CFL number: 0.16E+01 at (125, 49, 25)
Max divergence: 0.43E+04 at ( 18, 26, 4)
Min divergence: -0.68E+04 at ( 19, 26, 3)
Max div. error: 0.83E+04 at ( 5, 72, 21)
Poisson Pert. : -0.11E+10
No. of Lagrangian Particles: 499713
Total Heat Release Rate: 3309.591 kW
Radiation Loss to Boundaries: 1184.308 kW

Any suggestions on how I might solve this problem? I would love to be
more specific but I can't so I welcome any suggestions no matter how
general they may be.

[rmcdermo]

It is pretty well impossible to diagnose the problem with only this
bit of information. Are you new to FDS? If so, I would recommend
starting with a simple problem so that you are confident you have the
problem set up correctly. Make sure you have read the User Guide.

http://fire.nist.gov/fds/documentation.html

If you think this is a problem with the code, then please simplify the
case as much as possible (i.e., try to recreate the problem with a
case that runs in seconds, literally, and get rid of ANY irrelevant
components of the input file: DEVCs, OBSTs, etc.) and submit your case
to the Issue Tracker.

http://code.google.com/p/fds-smv/issues/list

[CM]

No, I'm not new to FDS.   I ran this same scenario in FDS 4 a while ago and it didn't give me instability problems. I was asked to rerun it with the updated version but so much changed between 4 and 5 that a lot of the input data had to be modified.  The jump in heat release rates seems to be the problem but it wasn't in FDS4.  I'll have to see how much a jump it was in 4 as compared to 5.

Unfortunately, I can't release much more information than this...but I was hoping somebody could maybe suggest stuff to look at...throw out even generalizations.  Long shot....

Oh well.  Thanks anyway.

[rmcdermo]

Here is my only "long shot" suggestion... In FDS5 5 the stability
check based on divergence (huge HRR leads to a big positive
divergence, i.e., expansion) is not invoked unless you specify
CFL_VELOCITY_NORM>0 (FDS 6 options). So, the first thing I always try
when I run into weird stability problems is set CFL_VELOCITY_NORM=1
and FLUX_LIMITER=2 on MISC. Hopefully this helps...

[CM]

Thanks for the suggestion.  I'll try that now and see what it gives me in the morning.

[JWilliamson]

I would also suggest that you do a test with a simple burner instead
of a propagating fire. You should also check the geometry at or near
the cells where the instabilities are highest to see if you have made
an error in input.

[pxlbarrel]

I thought I'd update this group about my numerical instability
problem.

I used the FDS6 options to see if it would solve the problem.
However, it didn't. It was, as mentioned earlier, a long shot. I've
tried other little tweaks including simplifying a bit but I'm still
getting the error whenever the heat release makes the jump. I've run
it on 3 different computers using 3 different platforms. The geometry
is rather uncomplicated and looking at it, it does not appear to be
distorted, stretched or otherwise compromised.

I have noticed that the divergence error seems to occur at different
mesh locations even when the same FDS file is used. I'm not sure if
that's significant in itself.

I'm at the point where I'm going simplify it to just having a large t-
squred fire and a small t-sqared fire so that I can get a range of
oxygen depletion times. I'll forego trying to recreate the flame
spread (as suggested earlirer) and hope it works.



On Sep 22, 4:23 pm, JWilliamson <williamson.justin.w...@gmail.com>
wrote:

> > > > general they may be.- Hide quoted text -
>
> - Show quoted text -

[Kevin]

It is time that this was posted to the Issue Tracker. We'll have to
run the case to diagnose it properly.

> > - Show quoted text -- Hide quoted text -