search

firemodels / fds

Fire Dynamics Simulator
https://pages.nist.gov/fds-smv/
Other
664 stars 623 forks source link

model gets numerical instability error #2274

Closed gforney closed 9 years ago

gforney commented 9 years ago 
Please complete the following lines...

FDS Version:5.5.3
SVN Revision Number:7031
Compile Date:29.10.2010
Smokeview Version/Revision:6.0.11
Operating System:win2008 server

Describe details of the issue below:
This is a large model that includes 4 tunnels [2 couples- north and south] and changing
slopes
the main part of the tunnels was handeled by GVEC command for simulating the slope
and the ramps that lead out from the tunnel were inclined by geometry.

1. using FDS6 I have received a 157 access violation bug
2. using FDS5, the run gets stuck
a. I tried fo split the model to parts and in 6 deifferent/short parts the model didn't
get stuck
b. splitting the model to 3 different parts/runs, the model gets stuck again
c. lowering resolution, taking off objects (like signs), didn't solve the problem as
well
d. the same for lowering volume flux.

Attached is the file that consists of the 2 northern tunnels only
I have been working on this model for months now
I will appreciate any assistance

Original issue reported on code.google.com by orit13a on 2014-12-29 12:57:40

gforney commented 9 years ago 
When I try to run the case with FDS 6.1.2, I get the following message:

ERROR: SURF_ID map1 cannot attach velocity boundary to thin obstruction (CHID: 9_116north)

This is a new restriction in FDS -- you cannot specify a velocity at the face of a
thin obstruction. Rework your input file, and run it with FDS 6.1.2, the latest release.
If there are still problems, submit the new input file. Try as much as possible to
simplify the case.

Original issue reported on code.google.com by mcgratta on 2014-12-29 14:43:29

gforney commented 9 years ago 
thank you
It is running
I will update whether it gets errors or not
Orit

Original issue reported on code.google.com by orit13a on 2014-12-31 09:15:57

gforney commented 9 years ago 
Hello,
The model runs slowly [fds6] but didn't get stuck yet.... it is at 75 secs now
attach is the command i used for Jet Fans and a picture
It seems as if the vectors go in both directions
am I wrong in my impression ?
did I have to write the command differently ?

&SURF ID='map3',
RGB=26,128,26,
TAU_V=5.0
VOLUME_FLUX=38.0/

&OBST XB=-124.0,-124.0,40.0,41.0,6.0,7.5, SURF_ID='map3' DEVC_ID='TIMER1'/ M3.29

I appreciate your assistance

Orit

Original issue reported on code.google.com by orit13a on 2015-01-07 13:45:30

gforney commented 9 years ago 
I cannot tell from your picture if the flow is going one way or the other. Look at a
slice of U only, not the vectors.

Original issue reported on code.google.com by mcgratta on 2015-01-07 13:52:49

gforney commented 9 years ago 
HERE IT IS

Original issue reported on code.google.com by orit13a on 2015-01-07 16:48:20

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

Original issue reported on code.google.com by orit13a on 2015-01-07 16:51:25

gforney commented 9 years ago 
You are assigning the same velocity to both sides of a thin obstruction. I am surprised
that the code even runs. Specify the SURF line on only one side of a thick obstruction.

Original issue reported on code.google.com by mcgratta on 2015-01-07 18:03:39

gforney commented 9 years ago 
I could not tell either
That is the problem
This is the u slice as you suggested
Since it is going so slow
I didnt want the simulation to keep going if its wrong

What can i do?

Original issue reported on code.google.com by orit13a on 2015-01-07 18:17:13

gforney commented 9 years ago 
I told you in Comment 7 what to do.

Original issue reported on code.google.com by mcgratta on 2015-01-07 18:49:38

gforney commented 9 years ago 
Dear Sir,

1. Could you please assist in pointing where exactly have I assigned the same velocity
to both sides of a thick obstruction.
The obstructions look very similar but the  are not adjective. [RELEVANT COMMAND LINES
ATTACHED AT THE END OF THIS MESSAGE]

2. After the model did run in 6 short parts, it still gets numerical instability when
all parts are put together, I have no clue as of how to solve the model problem.

I appreciate your assistance
Orit

&SURF ID='map1',
      RGB=26,128,26,
      VOLUME_FLUX=38.20,
      TAU_V=5.0/
&SURF ID='map2',
      RGB=26,128,26,
      VOLUME_FLUX=-38.20,
      TAU_V=5.0/

&OBST XB=-1377.0,-1377.0,48.0,49.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER'/ M2.7  
33
&OBST XB=-1377.0,-1377.0,46.0,47.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER'/ M2.8  
34

&OBST XB=-1306.0,-1306.0,46.5,48.0,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER'/ M2.8  
  32
&OBST XB=-1306.0,-1306.0,48.5,50.0,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER'/ M2.7  
31

&OBST XB=-1225.0,-1225.0,50.0,51.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER'/ M2.7  
30
&OBST XB=-1225.0,-1225.0,48.0,49.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER'/ M2.8  
 29

&OBST XB=-1124.0,-1124.0,51.0,52.5,5.25,6.75, SURF_ID='map1' DEVC_ID='TIMER'/ M2.7
   28
&OBST XB=-1124.0,-1124.0,49.0,50.5,5.25,6.75, SURF_ID='map1' DEVC_ID='TIMER'/ M2.8
     27

&OBST XB=-155.0,-155.0,52.5,54.0,6.5,8.0, SURF_ID='map2' DEVC_ID='TIMER'/ M2.1    
6  בכיוון T2 
&OBST XB=-155.0,-155.0,50.5,52.0,6.0,7.5, SURF_ID='map2' DEVC_ID='TIMER'/ M2.3    5
&OBST XB=-155.0,-155.0,54.5,56.0,6.0,7.5, SURF_ID='map2' DEVC_ID='TIMER'/ M2.2   4

&OBST XB=-53.0,-53.0,52.0,53.5,6.0,7.5, SURF_ID='map2' DEVC_ID='TIMER'/ M2.2    3
&OBST XB=-53.0,-53.0,48.0,49.5,6.0,7.5, SURF_ID='map2' DEVC_ID='TIMER'/ M2.3    2
&OBST XB=-53.0,-53.0,50.0,51.5,6.5,8.0, SURF_ID='map2' DEVC_ID='TIMER'/ M2.1    1 

&OBST XB=-1377.0,-1377.0,17.0,18.5,9.0,10.5, SURF_ID='map1' DEVC_ID='TIMER14'/ M2.8
   1
&OBST XB=-1377.0,-1377.0,19.0,20.5,9.0,10.5, SURF_ID='map1' DEVC_ID='TIMER14'/ M2.7
  2

&OBST XB=-1304.0,-1304.0,20.0,21.5,6.5,8.0, SURF_ID='map1' DEVC_ID='TIMER13'/ M2.7
  3
&OBST XB=-1304.0,-1304.0,18.0,19.5,6.5,8.0, SURF_ID='map1' DEVC_ID='TIMER13'/ M2.8
   4

&OBST XB=-1226.0,-1226.0,20.0,21.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER12'/ M2.8
   5
&OBST XB=-1226.0,-1226.0,22.0,23.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER12'/ M2.7
  6

OBST XB=-1097.0,-1097.0,20.0,21.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER12'/ M2.8 
  7 לא עובד כי ליד שריפה
OBST XB=-1096.46,-1096.46,21.9377,23.4372,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER12'/
M2.7 8

&OBST XB=-1025.49,-1025.49,18.453,19.952,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER11'/
M2.8 9
&OBST XB=-1025.0,-1025.0,20.5,22.0,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER11'/ M2.7
   10
&OBST XB=-931.0,-931.0,19.0,20.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER10'/ M2.8  
   12
&OBST XB=-931.0,-931.0,21.0,22.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER10'/ M2.7  
 11

&OBST XB=-788.0,-788.0,19.0,20.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER9'/ M2.8   
14
&OBST XB=-788.0,-788.0,21.0,22.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER9'/ M2.7   
 13

&OBST XB=-711.0,-711.0,20.5,22.0,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER8'/ M2.7  
    15  
&OBST XB=-712.0,-712.0,18.5,20.0,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER8'/ M2.8  
16   

&OBST XB=-639.0,-639.0,21.0,22.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER7'/ M2.7  
18
&OBST XB=-639.0,-639.0,19.0,20.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER7'/ M2.8  
17

&OBST XB=-536.0,-536.0,21.0,22.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER6'/ M2.7  20
&OBST XB=-536.0,-536.0,19.0,20.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER6'/ M2.8  19

&OBST XB=-445.0,-445.0,17.0,18.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER5'/ M2.8  
 22
&OBST XB=-444.0,-444.0,19.0,20.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER5'/ M2.7  
 21
&OBST XB=-362.0,-362.0,15.0,16.5,6.0,7.5, SURF_ID='map1', DEVC_ID='TIMER4'/ M2.8  
24
&OBST XB=-362.0,-362.0,17.0,18.5,6.0,7.5, SURF_ID='map1', DEVC_ID='TIMER4'/ M2.7  
23
&OBST XB=-286.0,-286.0,13.5,15.0,6.0,7.5, SURF_ID='map1', DEVC_ID='TIMER3'/ M2.8  
25
&OBST XB=-286.0,-286.0,15.5,17.0,6.0,7.5, SURF_ID='map1', DEVC_ID='TIMER3'/ M2.7  
26
&OBST XB=-200.0,-200.0,15.0,16.5,6.0,7.5, SURF_ID='map1', DEVC_ID='TIMER2'/ M2.7  
28
&OBST XB=-200.0,-200.0,13.0,14.5,6.0,7.5, SURF_ID='map1', DEVC_ID='TIMER2'/ M2.8  
27

&OBST XB=-125.0,-125.0,19.0,20.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER1'/ M2.7  
 29
&OBST XB=-125.0,-125.0,17.0,18.5,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER1'/ M2.8  
30
&OBST XB=-39.0,-39.0,17.5,19.0,5.5,7.0, SURF_ID='map1', DEVC_ID='TIMER'/ M2.8   31

&OBST XB=-39.0,-39.0,19.5,21.0,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER', DEVC_ID='TIMER6'/
M2.7   32
&OBST XB=-39.0,-39.0,4.5,6.0,10.0,11.5, SURF_ID='map1' DEVC_ID='TIMER'/ M2.8    33

Original issue reported on code.google.com by orit13a on 2015-01-11 09:05:35

gforney commented 9 years ago 
Start with your first OBST line:

&OBST XB=-1377.0,-1377.0,48.0,49.5,5.5,7.0, SURF_ID='map1' DEVC_ID='TIMER'/ M2.7  
33

The SURF 'map1' is applied to both the left and right (plus/minus x direction) side
of this obstruction. This is not the proper way to assign a velocity or volume flow
boundary condition. You must create an obstruction that is at least one cell thick,
and you must apply the SURF to only one side, using either SURF_IDS or SURF_ID6.

Original issue reported on code.google.com by mcgratta on 2015-01-11 16:22:53

gforney commented 9 years ago 
Thank you, I will do that

Original issue reported on code.google.com by orit13a on 2015-01-12 07:01:10

gforney commented 9 years ago 
1. I did as you suggested but the problem of directions still exists

2. "Regardless of the orientation of the plane x = 5,
the flow will be directed into the room because of the sign of VEL"

reading this......
I am now thinking
how does the code know which way do I want the jet fan to blow in a tunnel that is
open on 2 directions ?

3. I have also changed the obstacles to vents and it sill doesn't work

**This is regardless to the fact that in FDS5 THE MODEL DID RUN IN 6 SHORTER PARTS
which means that after I solve the FDS6 vent issue I might still have the model problem.

DO YOU HAVE A SUGGESTION ?
Will jet fan design with HVAC be better in your opinion ?

Thank you
Orit

Original issue reported on code.google.com by orit13a on 2015-01-13 14:15:36

gforney commented 9 years ago 
I suggest that you create a very small test case. A simple square duct and a single
jet fan. Work with this until you understand the directional convention, or if you
have problems, post the simple input file. The file should be no longer than about
20 lines.

Original issue reported on code.google.com by mcgratta on 2015-01-13 15:13:50

gforney commented 9 years ago 
Dear Sir
In the small model everything works just fine
In the large model after recreating all the Jet Fans in the tunnels (about 60 of them)
with HVAC after testing HVAC method in a small file in order to be sure I understand
how it works

I receive now an access violation [157]
what can the problem be now ?

Thank you very much

Original issue reported on code.google.com by orit13a on 2015-01-26 11:55:18

gforney commented 9 years ago 
I cannot read the .rar file. Just upload the input file in its native, plain text format.

Original issue reported on code.google.com by mcgratta on 2015-01-26 13:59:45

gforney commented 9 years ago 
It didnt get through
בתאריך 26 בינו 2015 15:59, <fds-smv@googlecode.com> כתב:

Original issue reported on code.google.com by orit13a on 2015-01-26 14:03:23

gforney commented 9 years ago 
How big is the .fds (input) file?

Original issue reported on code.google.com by mcgratta on 2015-01-26 14:11:53

gforney commented 9 years ago 
15.4mb
בתאריך 26 בינו 2015 16:12, <fds-smv@googlecode.com> כתב:

Original issue reported on code.google.com by orit13a on 2015-01-26 14:31:56

gforney commented 9 years ago 
I cannot debug a 15 MB input file. Are you using FDS 6.1.2?

Original issue reported on code.google.com by mcgratta on 2015-01-26 14:36:21

gforney commented 9 years ago 
Yes...
 בתאריך 26 בינו 2015 16:36, <fds-smv@googlecode.com> כתב:

Original issue reported on code.google.com by orit13a on 2015-01-26 14:43:57

gforney commented 9 years ago 
checked again, i was using FDS 6.0.11
do you think the new version will solve the 'access violation' ?
I will try that in the morning

Original issue reported on code.google.com by orit13a on 2015-01-26 21:02:09

gforney commented 9 years ago 
I assume you mean 6.0.1. Try the latest, 6.1.2.

Original issue reported on code.google.com by mcgratta on 2015-01-26 21:42:10

gforney commented 9 years ago 
Dear Sir
Trying to run the case with FDS 6.1.2
still gives access violation

Original issue reported on code.google.com by orit13a on 2015-01-27 10:04:55

gforney commented 9 years ago 
I am having trouble running this case too. I looked at the geometry and it is very complex.
Have you run any case successfully with this geometry? I would not recommend this approach
simply because you are only using about 10% of the domain -- the rest is outside the
tunnel and of little value. Maybe you should try another approach.

I will work on the case because it is important that FDS be able to handle large domains,
but I would urge you to consider an alternative approach.

Original issue reported on code.google.com by mcgratta on 2015-01-27 14:27:05

gforney commented 9 years ago 
Thank you fot the quick response
The thing is that i have already ran many tunnel cases before. This is my
office's speciality.
I did try to reduce to minimum the outside area and most of the inside area
will be in use once the fire spreads.
This full model got stuck with fds5. But did not get stuck with fds5 when
splitt to six parts.

Trying to solve this numerical instability I ran the case with fds6. most
updated version...

What do you mean by alternative approach.
I have a contract for fds for the project and can not supply the results.

Thanks again
Orit
בתאריך 27 בינו 2015 16:27, <fds-smv@googlecode.com> כתב:

Original issue reported on code.google.com by orit13a on 2015-01-27 15:22:58

gforney commented 9 years ago 
Orit,

As a test, try setting the gravity to a normal -Z value and try the run again.
I think that I have found a bug with gravity tables and multiple meshes.
I am working to confirm it, but it would be interesting to see if this single change
makes a difference for you.

-Bryan

Original issue reported on code.google.com by klein@thunderheadeng.com on 2015-01-28 15:02:30

gforney commented 9 years ago 
Thank you

I tried that

It gave the same error

Orit

Original issue reported on code.google.com by orit13a on 2015-01-29 10:15:21

gforney commented 9 years ago 
Orit, at the moment we cannot run your case here at NIST because we are having memory
limit issues on both of our linux clusters. What kind of machine are you running these
cases on?

Original issue reported on code.google.com by mcgratta on 2015-01-29 13:42:06

gforney commented 9 years ago 
Windows 2012 server 64 bit
בתאריך 29 בינו 2015 15:42, <fds-smv@googlecode.com> כתב:

Original issue reported on code.google.com by orit13a on 2015-02-01 15:29:32

gforney commented 9 years ago 
I have narrowed down the problem. Between Mesh 1 and Mesh 11 there is a problem. See
the attached imaged. The FDS_Grid shows the intersection of Meshes 1 and 11. The User_Grid
shows what you and/or PyroSim generated. You can toggle between these two views in
Smokeview by hitting the q key. Notice that when FDS "snaps" the geometry to the specified
grid, there are holes and misalignments. The original user-specified geometry is relatively
clean. By default, FDS sets up heat transfer in the solid obstructions such that heat
can be transferred through an obstruction. If the obstruction falls at a mesh boundary,
arrays have to be set up to transfer this data between computers. In your case, something
is going wrong here. 

I suggest that you generate a geometry on the same mesh that you plan to use for the
simulation. Your current geometry does not conform very well to the grid, in which
case there are significant changes when FDS shifts the solids to align with the mesh.
I will try to fix the bug in FDS, but I suggest that you clean up this geometry and
not risk having your geometry changed significantly by FDS.

Original issue reported on code.google.com by mcgratta on 2015-02-02 20:39:59

gforney commented 9 years ago 
I have more information. Do not use the THICKEN feature in a case like this. These thickened
obstructions cross mesh boundaries and cause confusion in the parallel calculation.
Since you are using PyroSim, you should design your obstructions to be exactly as you
think they should be. Do not use THICKEN in this case.

I identified a few bugs as well, which I fixed. But for the moment, try your case with
no THICKEN attributes.

Original issue reported on code.google.com by mcgratta on 2015-02-03 22:11:49

gforney commented 9 years ago 
Kevin
I have a question
since we have reached almost a dead end now [with FDS6] because the geometry of the
model is so complicated and took days to acheive.
And since the problem was smaller with FDS5, because the case DID RUN but had a numerical
instability in the full version, and DIDN'T GET STUCK when splitt to 6 parts.
Is there an option to try and solve this problem with FDS5 and not FDS6 ??

If this is possible I will upload the relevant file again

Orit

Original issue reported on code.google.com by orit13a on 2015-02-19 13:21:28

gforney commented 9 years ago 
If you are asking me to try to fix this case using FDS 5, the answer is no. The only
thing I can recommend to you is to set up your geometry with a mesh that is exactly
the same as the one you will use in the simulation. Do not use THICKEN either. In Smokeview,
when you hit the "q" button, you should see no shift in the obstructions. When I do
this with your current input file, I see things shift significantly. I see holes open
up, and I see these small passageways change dramatically. FDS can only do so much
to repair the damage.

Original issue reported on code.google.com by mcgratta on 2015-02-19 13:32:13

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

Original issue reported on code.google.com by mcgratta on 2015-04-29 16:53:48