Issues with fire simulation in a tunnel

[Jerome Morrow]

Hello everybody,

I got some issues with a simulation of fire inside a tunnel, and although it is documented in the FDS guidebooks (chapter 9.3 - Pressure considerations in long tunnels), there is still some details that I don't really understand and I'm not sur how to solve them:

I ran my simulation with FDS 6.7.4, and it stopped after around 510 s with the message forrtl: severe (157): Program Exception - access violation (before that, everything semmed ok). I tried on an other computer, same result.

After some researches on this group, I tried to run it with the latest version (6.7.9). After around 400 s, it stopped due tu nomerical instability, with some warnings about minimum density in the previous steps.

I modified the code, adding TUNNEL_PRECONDITIONER=T and setting MAX_PRESSURE_ITERATIONS to 50, as suggested in the FDS guidebook. It is running now, we will see if it still crashes.

Why did my first simulation crashed? And why did I get a numerical instability with the latest version, which did not occur with the older one? Anything else that I should do to avoid these errors?

Please find attached the code.

Best regards.

[Jerome Morrow]

With the last modification, it stops after only 24 s (numerical instability).

What should I do? Create some holes on the floor of the tunnel?

[fde]

Not an answer but a note: Extraction SURF is applied to many VENTs. But be aware that 23.33 m3/s is applied to each individually. For example, in the smallest VENT of 2.16 m2 area, you might observe velocity higher than 10 m/s.

[Randy McDermott]

I'm running the case now.

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/e5488ffc-2160-4383-b8a4-6b5acfe2b92bn%40googlegroups.com.

[dr_jfloyd]

The numerical instability error message happens when FDS determines that the timestep has been reduced to much compared to the initial timestep, this check occurs immeadiately after the next timestep size is determined.

Access violation means that FDS is trying to access an array outside of its bounds. A common reason is when some portion of the code tries to get a gas thermophysical property like specific heat outside the range of temperatures that FDS tabulates values for (default range is 0 to 5000 K). For a typical fire scenario getting a temperature outside this range means something has gone wrong in the simulation. This could be a bug somewhere resulting in a bad temperature value, but it can also be due to an instability developing where the instability first manifests as a bad temperature before the timestep adjustment.

The instability stop is an all or nothing check. Each new version of FDS comes with changes to the calculations and bug fixes. Fire is chaotic and changes to the code can mean changes to the timestep size during the calculation. This can mean that one version just barely crosses over the threshold for an instability stop but another version doesn't.

[Kevin McGrattan]

This case is similar to the one posted in https://groups.google.com/g/fds-smv/c/bJZjkMy2_I4

There are two MISC lines -- there should only be one.

[Jerome Morrow]

Thanks you all for your replies.

@fde Yes I know, but it is how the system is designed (210 m3/s passing through 9 vents of 2.16 m2).

@randy.m If you took the case as I uploaded it, there's no  TUNNEL_PRECONDITIONER=T as it is the first model I tested. Let me know what happen (it crashes after around 400 s in my case).

@dr_jfloyd Thank you for the explanations. What should I do to avoid these? As mentionned, I tried with  TUNNEL_PRECONDITIONER=T but it crashes rapidly. I am running now a new simulation with small holes on the floor but it is very slow...

@Kevin You're right. I corrected this and try to run it again with  TUNNEL_PRECONDITIONER=T, but it still crashes after a while (around 100 s).

[Kevin McGrattan]

I did not notice at first that you are using meshes of different heights. That probably is causing trouble for the TUNNEL_PRECONDITIONER because it computes the solution of the 1-D pressure equation and uses this solution to estimate the solution of the 3-D pressure equation. I am running the case now with all the meshes the same height. I added an OBSTruction to model the low ceiling portion of the tunnel. 

[Kevin McGrattan]

This case has run out to 60 s and appears to be OK for now.

[Randy McDermott]

The attached input file (no preconditioner, just tighter velocity tolerance and fixed MISC lines) is still running at 185 s with latest source.

On Wed, Jan 4, 2023 at 1:06 PM Kevin McGrattan <mcgr...@gmail.com> wrote:

This case has run out to 60 s and appears to be OK for now.

--

You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/CAAJimDF8dwQNCrnP9A_eBuRKD0ODZXZB9ckVBmur_QP19KS3cg%40mail.gmail.com.

[Jerome Morrow]

Thank you Kevin and Randy for your reply.

I am running it without TUNNEL_PRECONDITIONNER, but with small holes on the floor and it seems to work (it is around 950 s now), although it is very slow compared to the previous simulations.

Next time I will use  TUNNEL_PRECONDITIONER=T and keep the same mesh height and width.

Regards

[Kevin McGrattan]

The small holes can work, but they tend to create regions of high velocity which lower the time step and make the calculation run longer. Tight tolerances and the tunnel preconditioner are more helpful in maintaining a stable pressure field.

[Jerome Morrow]

Dear all,

I re-open this subject as I'm trying now to simulate an even longer tunnel, and it keeps crashing. I tried :

• TUNNEL_PRECONDITIONNER=.T. 
• MAX_PRESSURE_ITERATIONS=100
• VELOCITY_TOLERANCE=0.001

Everytime it is the same : Warnings about minimum density, and a few steps later numerical instability occurs.

I know I could create some small holes, but I know that my simulation will take forever with it...

Any other suggestions?

Regards,

[Kevin McGrattan]

I will run the case. At what time in the simulation does it fail?

[Jerome Morrow]

Thank you. It fails around 240 s.

[Kevin McGrattan]

OK, my run is out to about 172 s. I'll let it continue and see if I get similar errors. I am running with the latest source code.

[Kevin McGrattan]

The case ended with a numerical instability at about 240 s with the latest source code. I tried some other variations over the weekend, but still no luck. I will keep trying. Your case is not too complicated, so I'll use this as a good test case.

[Kevin]

This thread is now an issue. Continue all discussion there.

https://github.com/firemodels/fds/issues/11428