search

firemodels / fds

Fire Dynamics Simulator
https://pages.nist.gov/fds-smv/
Other
650 stars 618 forks source link

FDS+Evac: people "froze" #1396

Closed gforney closed 9 years ago

gforney commented 9 years ago 
Application Version: FDS 6.0.0 / FDS+Evac 2.4.1
SVN Revision Number: 8083 / 7989
Compile Date: compiled with GNU (gfortran)
Operating System: Win XP 32bit

Application Version: FDS 6.0.0 / FDS+Evac 2.4.1
SVN Revision Number: 8113 / 7989
Compile Date: compiled with Intel
Operating System: Win XP 64bit

for Timo:
I "played" with FDS+Evac 2.4.1 in a simple test case study. After a series of calculations
I observed that in one calculation of the calculation series (about 20 simulations)
one agent froze:

testcase001_korrektur4:
agent No 104 (x=1.98, y=9.97, z=1.5, v_0=1.18)

testcase001_korrektur5:
agent No 835 (x=19.67, y=13.37, z=1.5, v_0=0.92)

testcase001_korrektur5a:
agent No 171 (x=18.58, y=15.4, z=1.5, v_0=1.0)

I can't explain why or find the mistake.

Gregor

Original issue reported on code.google.com by gregor.jaeger on 2011-06-06 20:30:25

gforney commented 9 years ago 
the last two files

Original issue reported on code.google.com by gregor.jaeger on 2011-06-06 20:31:29

gforney commented 9 years ago 
(No text was entered with this change)

Original issue reported on code.google.com by mcgratta on 2011-06-06 20:50:55

gforney commented 9 years ago 
I'll look this.

But just a short check list for "froze":

1) How wide are your pathways? If there are smaller
   holes than about 65cm, then the largest agents
   might get stuck. Largest male body diameter is
   close to 60 cm and then you have a body-wall
   repulsion, so some agent might get stuck on a
   hole that is only 60 cm wide.

2) The agents go towards walls and corner pockets if
   you (main) evacuation mesh is not forming a single
   evacuation zone ("zone" is like the pressure zones
   in the fire part, i.e., one zone is connected, there
   is a way from one grid cell to every other grid cell
   that is not a solid. There should not be any door=>door
   or door=>entr between the grid cells in the same evacuation
   zone. (door=>entr/door is a solid wall.) This you can check
   by putting the SLCF PBZ=xx, VECTOR=.TRUE., QUANTITY='VELOCITY',
   EVACUATION=.TRUE./ so you can see the guiding flow vectors
   in Smokeview. If thye are going towards walls then you know
   that there is something wrong with our "evacuation zone".
   If you have many evacuation zones in a one floor of the building,
   you could make main evacuation meshes for each of these zones.

But I'll look the case.

TimoK

Original issue reported on code.google.com by tkorhon1 on 2011-06-07 07:54:35

gforney commented 9 years ago 
I checked this thinks before I posted this issue. In the flow vector field I can't find
any inaccuracy. My pathways are wide (simple room: 27mx 27m, door wide: 5 m) an very
simple. 

I'm looking forward to the solution.

Gregor

Original issue reported on code.google.com by gregor.jaeger on 2011-06-07 08:27:03

gforney commented 9 years ago 
Oh, now is see. I did not have any problems with:

Fire Dynamics Simulator
Compilation Date : Mon, 06 Jun 2011
Version: 6.0.0; MPI Disabled; OpenMP Disabled
SVN Revision No. : 8453

The problem with your SVN version is that it has some
falling down model and in your SVN version it was probably
(accidentally) used by default. I noticed this sometimes and
made the things so that the falling down model is not as
the default. If you have your results, see the color of
the "frozen" agent. It should be changed to cyan, when
it frozes. Well, COLOR_METHOD=0, so you should see this
change if you use some avatar in Smokeview that can use
this information (disk, ellipsoid, human_altered_with_data)
and show/hide => humans => color is chosen.

Put next things (some or all, I do not know):

TAU_FALL_DOWN= -1.0   (or any negative real number)

If the above does not help, then:

D_OVERLAP_FALL= 10000.0 (some large number, much larger than body size)

F_MIN_FALL = 9000000000000.0_EB
F_MAX_FALL = 9000000000000.0_EB
(Really large forces, so nobody falls down)

It might be that TAU_FALL_DOWN < 0 will do the job.
It depends on the SVN version. 

For your information: The fall down model is a really
simple one. It checks if the agents overlap too much
(and/or) if the resultant contact force is too large.
The tau_fall_down is the probability that an agent
falls down in a time unit (1/s) if the overlap and/or
force criteria are filled. This model is just for
academic interest at this point. It could be done
much better but first I'll see how this works for
some real cases. If it seems to be qualitatively
nice then that's enough.

Check if this helps and reply, so I could close this issue.

WBR,
TimoK

Original issue reported on code.google.com by tkorhon1 on 2011-06-07 10:00:11

gforney commented 9 years ago 
Thanks, I will check it in the evening and reply. 

Gregor

Original issue reported on code.google.com by gregor.jaeger on 2011-06-07 10:06:11

gforney commented 9 years ago 
Application Version: FDS 6.0.0 / FDS+Evac 2.4.1
SVN Revision Number: 8453 / 8200
Compile Date: compiled with GNU (gfortran)
Operating System: Win XP 32bit

After I compiled a new version, my test works fine. An other example (e.g. lecture
hall) works also fine. Thanks for your information. You can close the issue.

Gregor

Original issue reported on code.google.com by gregor.jaeger on 2011-06-07 20:01:13

gforney commented 9 years ago 
Thanks Gregor, I'll close this issue now.

Timo

Original issue reported on code.google.com by tkorhon1 on 2011-06-08 07:19:44