Fix BCs and geometry for pinball

[jcallaham]

Comparing with the recent TCFD article describing this flow configuration, this PR makes the following changes:

• Sign error in Dirichlet BC for solid-body rotation
• Increase actuator limits on Pinball
• Decrease "tau" value for first-order filter on Pinball
• Fix the cylinder locations (x-locations were slightly wrong)
• Rename the "fine" mesh to "medium" and make this the default
• Add a new "fine" mesh with domain extent matching the paper (but with around 3x more total mesh elements)

The paper also uses a prescribed velocity of (1.0, 0.0) for the free-stream boundary conditions, but I'm leaving as symmetry here. I doubt it makes much of a difference in the dynamics, and I expect symmetry BCs will work better on smaller domains.

C.C. @cl126162 @smokbel @ReHoss

[ReHoss]

Hello @jcallaham

Thank you for the update.

Comparing with the recent TCFD article describing this flow configuration, this PR makes the following changes:

• Sign error in Dirichlet BC for solid-body rotation

Is it something critical, or this does not substantially change results?

• Increase actuator limits on Pinball

I know that the implementation is inspired by the seminal papers on Pinball.
Why is this increase necessary? Is it solely to match the TCFD article?

• Decrease "tau" value for first-order filter on Pinball
• Fix the cylinder locations (x-locations were slightly wrong)

Wrong w.r.t. to the recent TCFD paper? But right with respect to the legacy papers?

I am totally fine with the update, I am just curious about its purpose:)

Thank you very much!
Best,

[jcallaham]

The actuator limits were increased to allow matching the values used in that article - they hadn't been chosen in a principled way before anyway. Unfortunately I would say that the other fixes could potentially change results for the actuated flow. Previously published papers didnt have as much detail, so basically the new one let me catch these errors, but they were probably incorrect to begin with. To be safe, I would try re-running anything with the pinball to be sure nothing changes dramatically. Sorry if that causes any headaches for you - let me know if you notice any weird behavior after this.

…

On Sat, Oct 5, 2024, 8:40 AM Hosseinkhan Rémy ***@***.***> wrote: Hello @jcallaham <https://github.com/jcallaham> Thank you for the update. Comparing with the recent TCFD article <https://link.springer.com/article/10.1007/s00162-024-00693-2> describing this flow configuration, this PR makes the following changes: - Sign error in Dirichlet BC for solid-body rotation Is it something critical, or this does not substantially change results? - Increase actuator limits on Pinball I know that the implementation is inspired by the seminal papers on Pinball. Why is this increase necessary? Is it solely to match the TCFD article? - Decrease "tau" value for first-order filter on Pinball - Fix the cylinder locations (x-locations were slightly wrong) Wrong w.r.t. to the recent TCFD paper? But right with respect to the legacy papers? I am totally fine with the update, I am just curious about its purpose:) Thank you very much! Best, — Reply to this email directly, view it on GitHub <#195 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AG6KRLTJO5C7ZR376AQE5TDZZ7M4DAVCNFSM6AAAAABPJVRIZCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGOJVGA2DIMZWHA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[ReHoss]

No worries! and thank you for the details, nothing critical from my side. I might come back with questions later on...

By the way, were you able to find a control solution with this sign error on the Pinball? Could this sign problem only affect the signal sign but not the information it contains? @jcallaham

Best,

[jcallaham]

Yes if you mean a stabilizing control, that was basically what precipitated finding these issues. Antisymmetric rotation of the two rear cylinders should now tend to stabilize the flow. If you use something like Re=100 with a control vector of [0, +6, -6] that should stabilize the wake as seen in the figures in that paper. Note that the paper uses a control term ("b") defined to match the linear velocity of the cylinder, not the angular velocity, so to reproduce those results in Hydrogym use a value of twice "b" from the paper.

…

On Sat, Oct 5, 2024 at 10:29 AM Hosseinkhan Rémy ***@***.***> wrote: No worries! and thank you for the details, nothing critical from my side. I might come back with questions later on... By the way, were you able to find a control solution with this sign error on the Pinball? Could this sign problem only affect the signal sign but not the information it contains? @jcallaham <https://github.com/jcallaham> Best, — Reply to this email directly, view it on GitHub <#195 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AG6KRLS57XXCRWAFKKL3AS3ZZ7ZVNAVCNFSM6AAAAABPJVRIZCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGOJVGA3TKMJXGY> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[ReHoss]

Hi @jcallaham ,

I have several questions or remarks regarding this update:

• Suppose that we ignore the slight coordinate change on the Pinball.
  This PR only impacts the steady flow computation of the Pinball and the RotaryCynlinder since it modifies self.bcu_actuation then their respective collect_bcu methods used by NewtonSolver.

• Before this PR, I computed some steady states and some snapshots of "natural flow" stepping the flows (Cylinder, Pinball and Cavity) with a zero actuation input (with the step method). Ignoring the fix on the boundary condition, the impact of the modification of TAU value should only modify actuated flows? In the case where zero actions are passed, this change should not have any effect on a (not actuated dynamics)?

• Does the RotaryCylinder object represent the setup where the actuation is a rotation of the physical cylinder while the Cylinder implements a blowing and suction setting?

• Regarding the actuator limits, for similar blowing and suction schemes we have now two different default maximum values (MAX_ACTUATOR = 10 for the pinball and MAX_ACTUATOR = 0.5 * np.pi for the cylinder).
  I think MAX_ACTUATOR should be equal for all environments (Cylinder and Pinball) with respect to their actuation mode (blowing and suction vs. rotation).

• Does the choice of MAX_ACTUATOR = 0.5 * np.pi for a rotary cylinder means the cyclinder rotates for at most 45 degrees in both directions ? Would increasing this value to 10 means the cylinder can rotate with an angle greater than $2\pi = 360$ degrees ?