Spurious results using composites with different discretization orders
******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello, I want to solve the incompressible flow around the naca0012 with nektar++. I have attached the input file that I am using for such a case. For nektar++ versions v5.1.0, v5.2.0 and 5.3.0 (master) some spurious results, two checkerboard patterns, around half and one chord length right above the trailing edge show up. Such a checkerboard pattern does not appear when using nektar++ v5.0.1 and v5.0.2. I have enclosed a screenshot showing the comparison between the different versions. The checkerboard pattern seems to first appear in the horizontal velocity, u. It develops right at the boundary between two composites with different discretization orders. This is not the only boundary between composites with different nummodes. The pattern does not belong to a transitory state, once it shows up it will not fade or get shed outside of the domain. It will eventually appear in the pressure and vertical velocity fields as well, but their magnitudes are considerably smaller compared to one present in the horizontal velocity. I have tried different approaches to mitigate it: - Increase and decrease nummodes. - SVV - Diferent precodnitiones for velocity and pressure - Different Reynolds number - Differnt operators for SmoothAdvection - Differnt operators for GlobalSysSoln The pattern only does not show up when using the same discretization order for all the composites. Any help would be welcome. Cheers, Guillermo
HI Guglielmo, Thanks for the feedback. It does look like there is a bug between the composite definition since it would look like not all degrees of freedom are being correctly constrainted. Can you send the mesh file along too so we can see if an ADRsolve also shows the same issue. Joao (cc’ed) has been looking at setting up a way to specify different polynomial orders in different regions and so this might be an alternative? Cheers, Spencer. From: nektar-users-bounces@imperial.ac.uk <nektar-users-bounces@imperial.ac.uk> on behalf of Guillermo Suarez <guillermo180395@gmail.com> Date: Tuesday, 31 January 2023 at 23:01 To: nektar-users <nektar-users@imperial.ac.uk> Subject: [Nektar-users] Spurious results using composites with different discretization orders ******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello, I want to solve the incompressible flow around the naca0012 with nektar++. I have attached the input file that I am using for such a case. For nektar++ versions v5.1.0, v5.2.0 and 5.3.0 (master) some spurious results, two checkerboard patterns, around half and one chord length right above the trailing edge show up. Such a checkerboard pattern does not appear when using nektar++ v5.0.1 and v5.0.2. I have enclosed a screenshot showing the comparison between the different versions. The checkerboard pattern seems to first appear in the horizontal velocity, u. It develops right at the boundary between two composites with different discretization orders. This is not the only boundary between composites with different nummodes. The pattern does not belong to a transitory state, once it shows up it will not fade or get shed outside of the domain. It will eventually appear in the pressure and vertical velocity fields as well, but their magnitudes are considerably smaller compared to one present in the horizontal velocity. I have tried different approaches to mitigate it: - Increase and decrease nummodes. - SVV - Diferent precodnitiones for velocity and pressure - Different Reynolds number - Differnt operators for SmoothAdvection - Differnt operators for GlobalSysSoln The pattern only does not show up when using the same discretization order for all the composites. Any help would be welcome. Cheers, Guillermo
Hi Spencer, Here is the mesh. Cheers, Guillermo On Sun, 2023-02-12 at 22:55 +0000, Sherwin, Spencer J wrote:
HI Guglielmo,
Thanks for the feedback. It does look like there is a bug between the composite definition since it would look like not all degrees of freedom are being correctly constrainted. Can you send the mesh file along too so we can see if an ADRsolve also shows the same issue.
Joao (cc’ed) has been looking at setting up a way to specify different polynomial orders in different regions and so this might be an alternative?
Cheers, Spencer.
From: nektar-users-bounces@imperial.ac.uk < nektar-users-bounces@imperial.ac.uk> on behalf of Guillermo Suarez < guillermo180395@gmail.com> Date: Tuesday, 31 January 2023 at 23:01 To: nektar-users <nektar-users@imperial.ac.uk> Subject: [Nektar-users] Spurious results using composites with different discretization orders
******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello,
I want to solve the incompressible flow around the naca0012 with nektar++. I have attached the input file that I am using for such a case. For nektar++ versions v5.1.0, v5.2.0 and 5.3.0 (master) some spurious results, two checkerboard patterns, around half and one chord length right above the trailing edge show up. Such a checkerboard pattern does not appear when using nektar++ v5.0.1 and v5.0.2. I have enclosed a screenshot showing the comparison between the different versions.
The checkerboard pattern seems to first appear in the horizontal velocity, u. It develops right at the boundary between two composites with different discretization orders. This is not the only boundary between composites with different nummodes. The pattern does not belong to a transitory state, once it shows up it will not fade or get shed outside of the domain. It will eventually appear in the pressure and vertical velocity fields as well, but their magnitudes are considerably smaller compared to one present in the horizontal velocity.
I have tried different approaches to mitigate it: - Increase and decrease nummodes. - SVV - Diferent precodnitiones for velocity and pressure - Different Reynolds number - Differnt operators for SmoothAdvection - Differnt operators for GlobalSysSoln
The pattern only does not show up when using the same discretization order for all the composites.
Any help would be welcome.
Cheers, Guillermo
participants (2)
-
Guillermo Suarez
-
Sherwin, Spencer J