Tunnel fire - good results with FDS5, goes unstable if FDS6=TRUE

[GoogleCodeExporter]

Please complete the following lines...

Application Version: 5.3.1 Serial
SVN Revision Number: 3881
Compile Date: 1 May 2009
Operating System: Windows XP

Describe details of the issue below:

I am running some simulations of a tunnel fire using FDS 5/6.  General 
parameters are: 

Tunnel 10m wide and 6.4m high, inlet velocity 2.6m/s.  Distance of 20 
hydraulic diameters allowed upstream of the fire.  5MW fire.  Fire 
geometry is confined to floor of tunnel, area of blockage is small and 
the fire sits on the tunnel centerline.  Symmetry (mirror) boundary 
applied on tunnel centreline (a bit dodgy for LES, but it has given 
reasonable results in the past). 

Reasonable results achieved with FDS5. 

With FDS6 turned on (FDS6 = TRUE) numerical instabilities occur. 
Trialled turning off the dynamic model and also reducing the heat 
release rate.  Again, numerical instabilities result.  When the fire 
boundary condition is not included in the model, no instability results, 
but when the fire is included (even if HRR = 0), the model goes unstable.

File is attached. 

Original issue reported on code.google.com by matt.bil...@gmail.com on 23 Aug 2009 at 11:29

Attachments:

• tv-8-13.fds

[GoogleCodeExporter]

We'll take a look at it. Have you run the case with BAROCLINIC=.FALSE.? 
BARO=.TRUE. 
can lead to numerical instabilities, which is why we don't make it the default. 
Thanks for looking into this.

Original comment by mcgra...@gmail.com on 23 Aug 2009 at 1:16

• Added labels: Milestone-FDS_6.0

[GoogleCodeExporter]

First, run with the latest version in the repository. We found major bugs in 
older
versions.  I am running your case now and all seems fine.  If your case still 
does
not work with the latest code, please give us a hint about when it fails.  And, 
as I
mentioned before, please simplify your input file.  Did you try halving your
resolution and do you still get an instability?  Are all those DEVC lines really
necessary for debugging this problem?  Please reduce the number of lines in the 
input
file as well as try to determine the coarsest resolution that still gives the 
same
instability.  This helps us greatly in diagnosing the problem.

But, like I said, please compile and run the latest SVN and see if you still 
have
problems.

Original comment by randy.mc...@gmail.com on 24 Aug 2009 at 4:21

[GoogleCodeExporter]

I am slowing chipping away at this.  I tried BAROCLINIC=.FALSE. and it seemed 
to 
work ok (got well past typical "crash" time of 50s).  I will download latest 
compiled SVN and try BAROCLINIC=.TRUE.

Thank you

Original comment by matt.bil...@gmail.com on 26 Aug 2009 at 10:54

[GoogleCodeExporter]

Matt,

With the latest SVN and FDS6=T, the case ran to 374 sec and then still 
encountered an
instability.  This problem seems to be nearly identical to one that Chris
Lautenberger is having with another case.  In his case, I tried running in 
debug mode
(which takes weeks) and it actually did not crash.  So, you can see how hard 
this is
to diagnose.  Please be patient.  I suppose the good news is that it is quite
possible these are due to the same problem.  My gut is telling me it has more 
to do
with DYNSMAG and the changes to the VELOCITY_BC routines than BARO.  I have a 
case
running with FDS5 and DYNSMAG=T that is up to 297.

More to come...
Randy

Original comment by randy.mc...@gmail.com on 26 Aug 2009 at 12:33

• Changed state: Started

[GoogleCodeExporter]

FDS5 (SVN 4590) with DYNSMAG=T works fine. Tonight I'll add the new FLUX_LIMITER
scheme and see how things fare.

Original comment by randy.mc...@gmail.com on 26 Aug 2009 at 8:27

[GoogleCodeExporter]

Attached case seems to be working with SVN 6493.  Please verify.

Original comment by randy.mc...@gmail.com on 15 Jul 2010 at 8:48

• Changed state: Fixed

Attachments:

• tv-8-13.fds

[GoogleCodeExporter]

Original comment by randy.mc...@gmail.com on 26 Aug 2011 at 7:47

• Changed state: Closed