Closed mcgratta closed 3 years ago
I added a WALL THICKNESS device and it looks like the error is happening when the pool fully evaporates. Probably not blowing but a bug relatead to the wall renoding.
How long do these take to crash? Any chance of bisecting? Last time I ran them they seemed to work. I agree we need these functional before a release.
A few minutes. Bisection is possible. I was just hoping that someone might recognize the problem and pinpoint the commit.
Is it just water evap that fails? The pool fires cases finished ok?
The VTT cases failed too, but they ran for days.
On Wed, May 5, 2021 at 12:38 PM Randy McDermott @.***> wrote:
Is it just water evap that fails? The pool fires cases finished ok?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/firemodels/fds/issues/9479#issuecomment-832839418, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACWPCF5ZV75VYQ5NUJOE76TTMFYADANCNFSM44DK5JPQ .
I have time to bisect this now. Is anyone else doing it?
Go ahead
On Wed, May 5, 2021 at 1:20 PM Kevin McGrattan @.***> wrote:
I have time to bisect this now. Is anyone else doing it?
— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/firemodels/fds/issues/9479#issuecomment-832869630, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABBWXQMIIATNOYTIO74JGG3TMF463ANCNFSM44DK5JPQ .
-- Sent from my iPhone
It is when the water layer disappears. If you shrink the water layer, it will die faster. Given when these last ran it is either the wall renoding changes in Feb. or the prevention of T_BOIL overshoot in Jan. The renoding changes passed all the pyrolysis verification changes and FAA and UMD polymers which leads me to suspect something in the routine that looks at stopping overshoot in the pool temperature. I've been at client site until a few minutes ag,o so haven't had much of a chance to dig into this.
I'm a few minutes away from nailing down when the problem started. Early January. I'll let you know.
efd9991f969e99c991d00d548abd3028e56ecebe is the first bad commit
commit efd9991f969e99c991d00d548abd3028e56ecebe
Author: drjfloyd <jfloyd@jensenhughes.com>
Date: Thu Jan 21 16:57:24 2021 -0500
FDS Source: Prevent overshoot of TMP_BOIL for a liquid wall cell
:040000 040000 3ebbb3a340610dbb9a1111df212950886c2d4332 5d59e3072eadea0c984bf8118b094092f155745a M Source
I see the problem. I am not checking correctly in one of the loops to see if the wall cell has the liquid. In the test cases I was using for this it was either one layer or the pool was backed by steel. In those cases we never had a wall cell over the boiling temperature that didn't contain liquid. In this case with a glass backing to the pool the glass gets hotter than boiling when the pool goes away and since I wasn't checking correctly it is trying to apply a mass loss correction to the glass.
That'll do it. Could you check the VTT input files and confirm if they fit this same description. They take days to run, but I want to make sure that they're having the same issue.
should be fixed now. The ASTM case runs now and matches the experiment quite well. VTT is heptane floating on water so I could see the issue happening there.
Can we take this opportunity to overhaul the VTT input files to use multiple meshes (to run overnight) and the name of the validation subdirectory to VTT_Pool_Fires?
Are all the cases in that directory from VTT? There is the ASTM water case, the two cases with VTT in the name, and five other cases.
Yes, all these cases basically belong to the paper that Simo and Topi did. I'd just like that to be "VTT_Pool_Fires". As Kevin said, besides the ASTM case, they take way too long to run. We just have never set up a multi-mesh input files for them. Now would be a good time. I can do this if you want. Just whoever is going to run them.
OK, I'm going to update the meshing scheme so that these cases run fast. But I'm going to hold off on renaming as this will take some days to make sure I get all the guides buttoned up. We can do that after release.
The 5 original Pool_Fires cases are just pulled from the SFPE Handbook chapter on Liquid Fuel Fires. The burning rates were just pulled from a chart relating burning rate and heat of gasification.
Yes, I know that. But the VTT pool fire cases are actually experiments. I just think it better describes that set of cases.
This all seems to be working.
I just ran the Pool_Fires validation cases and the ASTM water evaporation and VTT heptane cases all failed with numerical instabilities. Last successful run was Dec 17. This might have to do with BLOWING=T, or some other recent change to evap model.