Xiangyu-Hu / SPHinXsys

SPHinXsys provides C++ APIs for engineering simulation and optimization. It aims at complex systems driven by fluid, structure, multi-body dynamics and beyond. The multi-physics library is based on a unique and unified computational framework by which strong coupling has been achieved for all involved physics.
https://www.sphinxsys.org/
Apache License 2.0
259 stars 199 forks source link

Numerical dissipation (non-linear wave fsi) #319

Closed ncsalis closed 11 months ago

ncsalis commented 11 months ago

Results are significantly worsened when simulating the case with high particle resolution (dx=0.015 circa 8M particles). This did not happened before merge #274 (prevented free surface rising when simulating with low particle resolution). Maybe it's better to revert this change since also @Shuaihao-Zhang encountered problems after this.

issue

ncsalis commented 11 months ago

Snapshot of the wave-structure impact before #274 and after

issue

Xiangyu-Hu commented 11 months ago

It really increases the numerical dissipation of the simulation. We probably can decrease the numerical dissipation, as a first check. As the present limiting parameter is based on previous formulation.

ncsalis commented 11 months ago

I will check the limiter in the following days. I have found this issue also in the 2d_nlwfsi case, the numerical dissipation is quite high even at low particle resolutions.

ncsalis commented 11 months ago

324 is ineffective, with lower limiter the free surface elevation does not improve. Also the dynamics (wave phase) of the wave is significantly modified and this aspect is unchanged

eta

ncsalis commented 11 months ago

In the previous simulations there was a typo in the wave gauge position (1 meter downstream compared to the exp). The dynamics of the wave is good. I have checked the simulation at different merge (#229, #271 and #274). The numerical dissipation was always present, just hidden by the typo.

eta

Xiangyu-Hu commented 11 months ago

So the new formulation is OK?

ncsalis commented 11 months ago

Yes, the new formulation is ok

Xiangyu-Hu commented 11 months ago

For numerical dissipation, Yaru's work will be much helpful. https://arxiv.org/abs/2304.14865

Xiangyu-Hu commented 11 months ago

I made another small change, but the result is similar.

ncsalis commented 11 months ago

Thank you for the suggestion on Yaru's paper. I also was thinking that first order consistency should improve the result

ncsalis commented 11 months ago

The first order consistency correction works very well. Result is now marginally improved

wave_correct

ncsalis commented 11 months ago

With the KGC seems that there is low dissipation thus with higher wave height that leads to higher amplitude of the fsi. How increase dissipation? I have tried by modifying the alpha parameter in the WKGC formulation but it doesn't produce significant result.

wave__

Xiangyu-Hu commented 11 months ago

Dear Nicolo,

The overestimate is most probably due to the lack of turbulent dissipation, which is not included in our model.

I think that we need not to tune the numerical dissipation and such slight overestimate, in my opinion, is acceptable.

Best regards,

Xiangyu


From: Nicolò Salis @.***> Sent: Tuesday, June 13, 2023 8:48:50 AM To: Xiangyu-Hu/SPHinXsys Cc: Xiangyu Hu; Comment Subject: Re: [Xiangyu-Hu/SPHinXsys] High numerical dissipation (non-linear wave fsi) (Issue #319)

With the KGC seems that there is low dissipation thus with higher wave height that leads to higher amplitude of the fsi. How increase dissipation? I have tried by modifying the alpha parameter in the WKGC formulation but it doesn't produce significant result.

[wave__]https://user-images.githubusercontent.com/56539341/245376066-78b4fe44-d10f-4c20-b0d7-d6efb20d7d31.png

— Reply to this email directly, view it on GitHubhttps://github.com/Xiangyu-Hu/SPHinXsys/issues/319#issuecomment-1588650450, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AI5OR4TVK64WT4D2ZUBCSN3XLAENFANCNFSM6AAAAAAY4FGPMU. You are receiving this because you commented.Message ID: @.***>

ncsalis commented 11 months ago

Ok, thank you!