Noticable Difference in Flow between 6.7.7 and 6.7.8

[bwklein]

We were running our test suite for the next version of PyroSim and one of our Airflow test cases reported an issue with the measured flow velocity with the latest 6.7.8 version of FDS. We started digging into the issue to see why the flow was different and have created a set of simple test cases to demonstrate the difference.

FDS_files.zip

There are 3 cases, each with a comparison screenshot below. Each image shows the simulation results in 6.7.7 and 6.7.8 side-by-side at the same point in time. The most significant difference was in the first case below, where a supply vent is placed on a mesh boundary. The velocity data for the device is significantly different in this case. Note that the max value of the colorbar on the right side image was not manually set to 20 m/s, so while the color is similar, it is actually lower peak flow velocity by almost 5 m/s.

We then thought that this might be some effect from the vent on the mesh boundary, and decided to try it with the supply surface assigned to the face of an obstruction. You can see a difference in the velocity and width of the flow from the surface between the two versions and a significant difference in flow pattern when compared to the vent on the mesh boundary case above.

We then thought that maybe this was a multiple mesh issue, so we ran the same case above, but only with a single mesh. You see that there is a similar difference with a single mesh as with the multiple-mesh case. Note that the max value of the colorbar on the left side image was not manually set to 20 m/s, so while the color is similar, it is actually lower peak flow velocity by almost 5 m/s.

Having said/shown that, we are not sure which is a 'better' answer, when I looked in the verification cases for 'AirFlow' to see if there were changes from 6.7.7 to 6.7.8, it looks like there were changes to what was simulated and measured there as well, so there is not a direct comparison in the Verification data/document. Did something change in how velocities are assigned to surface cells in the latest release?

Thank you for taking a look and your advice on how to move forward.

[mcgratta]

My guess is that this is related to this issue. If in your case there is no obvious right or wrong, I would suggest we not change anything that would adversely affect these jet fans that are coarsely resolved.

[bwklein]

Why do you think in 6.7.8 there is such a significant difference between a vent on the mesh boundary and the same size face on an obstruction? I didn't see that issue described in the case that you sent me.

[drjfloyd]

The issue Kevin pointed to was a change in the behavior of tangential flow for an HVAC VENT. This resulted from inconsistent treatment of edges and corners that had crept into the code.

In your case with the VENT on the boundary, the surface is flat. In your case with the OBST, there is a raised edge.

[bwklein]

@drjfloyd I modified the Mesh Boundary case to put the same vent on a larger obstruction so that the edges of the vent were on a flat surface and I get a very similar flow to when it is placed on the mesh boundary directly. So this does confirm that the 'raised edges' are the key to the more significant difference in the results we are seeing. Having said that, I expect that people are going to get very different results now (lower velocities and more 'turbulent' flow from vents on flat surfaces), with the next maintenance release (6.7.8) when compared to many previous versions of FDS. I would suggest making a note about this difference in release notes for 6.7.8 so that if doesn't catch folks unaware.

I consider this issue to be explained, verified, and resolved. We will update our test cases with new 'expected' values.
Thanks all for digging in and sorting this out.

[rmcdermo]

Thanks Bryan. I missed adding this to the release notes.

On Mon, Jun 6, 2022 at 11:38 PM Bryan Klein ***@***.***> wrote: @drjfloyd <https://github.com/drjfloyd> I modified the Mesh Boundary case to put the same vent on a larger obstruction so that the edges of the vent were on a flat surface and I get a very similar flow to when it is placed on the mesh boundary directly. So this does confirm that the 'raised edges' are the key to the more significant difference in the results we are seeing. Having said that, I expect that people are going to get very different results now (lower velocities and more 'turbulent' flow from vents on flat surfaces), with the next maintenance release (6.7.8) when compared to many previous versions of FDS. I would suggest making a note about this difference in release notes for 6.7.8 so that if doesn't catch folks unaware. I consider this issue to be explained, verified, and resolved. We will update our test cases with new 'expected' values. Thanks all for digging in and sorting this out. — Reply to this email directly, view it on GitHub <#10655 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABBWXQMFUZLBCWZJUZZXKMTVNZVVXANCNFSM5YAAJDSQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Sent from my iPhone

[mcgratta]

A comment was added to the FDS Release Notes.