Constant Numerical Instability/Pressure Issue in Imported Model - FDS

[idearies]

Hello,

We're currently trying to run a fairly detailed imported model for a client using Pyrosim/FDS and have consistently encountered the numerical instability issue which consistently crashes the simulation. The model contains more unique infrastructure with non-uniform gable roofs and this is the first time we've constantly had this issue in a project. We were previously running on an older version of the model, but both models have the same issue which we think is because of the lack of make up air or exportation of CAD models to FDS generating lots of thin-walled zones and compartments leading to many pressure zones that end up giving Minimum density issues and instability crashes.

We've tried:

• Changing the cell size to 0.25, 0.5, 1.0
• Running separate zones and combined zones. We did find that we could run each zone on their own but once we had multiple zones running together or generally a bigger area the crash happens.
• Increasing openings to allow air to enter the building
• Enabling and disabling exhaust fans/adding fresh air blowers

Below is the link to the Pyrosim/FDS files. Please check Zones 2,4 + 6 Test for the fds and out file of the current model and the old model folder can be optionally check if needed.

Note that most things tested above were on the older model since we had just recently gotten the new version.

https://drive.google.com/drive/folders/1_c9axyZozjDQUisSSnciXk8D6i8ijayE?usp=sharing

We're lost on which direction to go further to get it to run smoothly because we aren't entirely sure how to diagnose and locate the areas actually causing the crash due to the possibilities and how time consuming each run or manually fixing parts can be. So any suggestions, leads, and identification of issues would be much appreciated.

Thank you

[drjfloyd]

The input file starts with
Generated by PyroSim 2024.2.1209

If you are running through Pyrosim then this is not the current release of FDS. If that is the case try with the current release. If the current release with the default solver doesn't work could try the current release with UGLMAT HYPRE

[mcgratta]

I will try to run this 1-_Gelephu_zone_2_4_fan_5_ach_20260108.fds with the latest version of the FDS source code. If it runs successfully, I am going to recommend that you upgrade. If it does not, it may be nearly impossible to determine why. The input file has 72000+ lines.

[mcgratta]

By the way, for this case, at what point in the simulation does the numerical instability occur?

[mcgratta]

The case is running but has made no progress thus far. There are 2717 "pressure zones"; that is, small, perfectly sealed pockets within the domain, most consisting of one or two cells. FDS still attempts to solve the flow equations within these pockets, which can lead to instabilities for a variety of reasons. My only advice is to simplify the geometry. The 3-D image looks impressive, but with all these pockets something is bound to go wrong.

Also, using an alternative pressure solver like UGLMAT HYPRE is not going to help because it would attempt to solve 2717 individual pressure equations.

[idearies]

Hello, thank you for your input. I will try to update the program tomorrow and run it with the latest source code as well, apologies for the oversight. For the latest file, the simulation time goes up to 224s out of 500 seconds at a time step of 3406. In other previous runs with the old model, depending on what we tried to change around and areas, it would usually cut off at around 80-150s.

Regarding your advice, is it safe to assume the most likely cause of the instability is due to the complexity of the model and thus the sheer amount of pressure zones that come with them? Are there any specific areas that are raising red flags such as the roof, glass, walls, etc. we might try to focus on simplifying, or is it best to just try to simplify it all that we can?

[mcgratta]

My guess is that the problem lies in how the CAD package renders the obstructions. You probably have many thin plates that form these pockets, and it would be much better to fill these pockets. Also, I don't know how you leakage and ventilation strategy works. For a model like this, the first thing I would do is just blow air into it and monitor the pressure. If you put a fire in a sealed space, the pressure is going to rise to unphysical values because $pV=nRT$.

[mcgratta]

The case failed with a numerical instability at around 253 s. You can see the location in the image. I cannot say what the cause of the problem is because I don't understand the geometry at this location.

[drjfloyd]

Your geometry is complicated; however, it has a lot of repeating elements like the trusses. It may not be as much work as you think to redraw large portions of this in Pryosim and greatly reduce the 1000s of pressure zones that resulted because CAD has exported each rectangular solid block as six planar faces. For example, a roof segment with multiple identical trusses under it. Draw the end truss as a series of rectangular solids in Pyrosim, group those into an object, and then use the Pyrosim move function to make copies at regular intervals below the roof.

[mcgratta]

Look closely at the image I posted. This image is a snapshot of the Plot3D file that was output when the case failed. It shows (via the large flow vectors) where the problem occurs.

[idearies]

I contacted my team and they said that in an older version of the model they have tried to redraw the roof segments before to fix the issue. However, the result was that a thin roof could run properly, but because the roof was so thin the smoke was leaking outside in a way that did not match our conditions. When the roof was thickened up, the numerical instability crash came back. Unfortunately I believe that the file for that version does not exist anymore since a lot of storage had to be cleared up since then, but we can try to continue to fill up the obstructions that we can in the new model in the meantime. I will have to follow up again with the project lead for any more specific details or additional inputs she can give to help out.

[tkorhon1]

Well, does the image show? It is a snapshot from the FDS User Guide: "10.3.4 Filling Pressure Zones". This usually helps me with these kinds of problems.

[mcgratta]

Anything further on this case?