About changes from FDS 6.6.0 to 6.7.9

[Hanzhou Mao]

I eventually managed to run jobs on FDS 6.6.0 (I tried adjusting my .fds file from FDS 6.7.9 to 6.6.0 as best as I could after numerous trials and errors).

However, it appears that there's some variability in the outputs, to say the least.

For example, for a given device measuring the mean value between 4 meshes :

On FDS 6.7.9 :

&DEVC ID='Cible', QUANTITY='ADIABATIC SURFACE TEMPERATURE', SPATIAL_STATISTIC='MEAN', XB=11.5,12.0,-15.0,-14.5,4.5,4.5/

On FDS 6.6.0 :

&DEVC ID='Cible', QUANTITY='ADIABATIC SURFACE TEMPERATURE', STATISTICS='MEAN', XB=11.5,12.0,-15.0,-14.5,4.5,4.5/

I would say the global structure, the materials and the reaction shouldn't have changed, yet there's a maximum value difference of 40°C in the output, which leaves me somewhat perplexed.

What big change between the 2 versions could I possibly have missed?

Thanks in advance,

HM

[dr_jfloyd]

There is 6 years of development and 1000s of source code commits between 6.6.0 and 6.7.9. With only knowing a DEVC input, there is no way to speculate why this might be the case.  You may want to review the FDS Release Notes and see if it mentions changes relevant to what you are modeling: https://github.com/firemodels/fds/wiki/FDS-Release-Notes

[Hanzhou Mao]

Thanks for your reply,

Yes, I've made changes based on what I saw on the release notes.

Some particular changes I've taken into account are :

• SIMULATION_MODE replaces RESEARCH_MODE to control the level of physics. For example, SIMULATION_MODE='DNS', sets the code to run in direct numerical simulation model. Other options include 'LES' (high-fidelity, convergent LES simulations), 'VLES' ("very large-eddy simulation" [default] for coarse-grained, practical LES calculations), and finally 'SVLES' ("Simple VLES", makes simplifying physical assumptions, such as constant specific heat ratio, to speed calculations when appropriate).

• Allow SPATIAL_STATISTIC for a linear array of devices.

• Add SPATIAL_STATISTIC to compute centroids.

So it appears VLES in set by default on 6.7.9, which was not the case on 6.6.0 (?), could it possibly be the root cause?

Plus, even when I tried using SPATIAL_STATISTIC on 1 mesh surface, the results are still a little different...

By the way, I've implemented a &SURF RAMP_Q, and didn't see changes about it in the notes between the 2 versions, so should I assume this part would work properly?

HM

[dr_jfloyd]

I have no idea what you are modeling and what 40 C difference means. I can't opine on what the cause might be. Are you seeing 40 C change that is 1000 C to 1040 C? This is not significant. Are you seeing 20 C to 60 C which is a large difference?  Are you specifying the heat release rate or modeling the heat release rate? There has been substantial development in the solid phase pyrolysis between 6.6.0 and 6.7.9. Do you have a multi-mesh case where there have been various bug fixes that could impact results? What is happening at that location in the domain and what is your grid resolution? If there is complex flow in that location, slight shifts in the location of say a plume or ceiling jet could move a hot spot in or out of the area you are computing the MEAN. There are too many possibilities to know what might be a reason with the information provided so far.

[Hanzhou Mao]

Are you seeing 20 C to 60 C which is a large difference? 

-> It's more like it, yes (100 to 140°C).

Are you specifying the heat release rate or modeling the heat release rate?

-> I've specified the HRRPUA value, and applied a custom RAMP_Q (the same on both versions).

Do you have a multi-mesh case where there have been various bug fixes that could impact results? What is happening at that location in the domain and what is your grid resolution? If there is complex flow in that location, slight shifts in the location of say a plume or ceiling jet could move a hot

-> I've seen this on the release notes, yes. In my case, none of devices nor the burning surface are present on overlapping meshes, so I assume it should be fine.

What is happening at that location in the domain and what is your grid resolution?

-> I've put my devices on the ceiling surface, with a 0.25x0.25x0.25 resolution (so a device should be measuring the mean value between 4 mesh surfaces, if everything is done correctly).

What's happening : hot gas is being accumulated in the ceiling (with some openings though) of a corridor, the device should be out of radiation impact range from the burning source.

 There are two many possibilities to know what might be a reason with the information provided so far.

-> Yes, sorry, I can only provide so much info.

[Hanzhou Mao]

So it appears on the plots that the difference in Tmax are mainly due to some fluctuations (see the green plot below : first one is 6.6.0, second one is 6.7.9).

To me the results in the 6.6.0 version look plausible (especially when the fluctuation are lessened), what would you think?

[dr_jfloyd]

The mean values are quite similar. Fluctuations are random events and very difficult to compare between versions, espeically with 6 years of development and 9 releases between the two versions. These two results are not what I would consider to be very different. You have no basis via which to decide which versions results are more plausible in this case. You don't have measured data for the specifiic fire scenario that tells you what the real answer is. Using 6.6.0 would mean that you are accepting to use a deprecated version of a software, instead of the current release that has fixed multiple bugs and made multiple improvements to the underlying numerics. 

[Randy McDermott]

In 6.7.6, we set the default near-wall turbulence model (model for the eddy viscosity of the first cell near the wall) to WALE.  I suspect this is the main reason for the differences in fluctuations you see.

--
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/03265f8d-0153-4d12-afe5-8ff351255c7an%40googlegroups.com.

[Hanzhou Mao]

Understood, then I'll go with these results.

Many thanks!

HM