Dear Shan,

Coupled NS, with ARPACK doesn't have a clear convergence history and gives you EV close to the pole. If you know the approximate position of the EV you can use Real and ImagShift to bring the pole closer to the EV you look for. It might be that this option only works with homogeneous expansion (?). Also you might increase the number of recovered EV and the size of the Krylov space used. For my cases I modified Nektar++ to calculate and return 1000+ EV (the hard coded limit is 256, I think), so this can be done.

Cheers!
Stan

20 lut 2022 03:39 shan xiangjun <x_j_shan@outlook.com> napisał(a):

Dear Stan,

 

I tried the CoupledLinearisedNS solver with ARPACK and got converged EV, although the structure of the eigenvectors is similar but the eigenvalues are quite different from the literature (e.g. unstable flow showing negative real part). EV also varies greatly with NumSteps.

 

Is there something wrong with the settings of my solver parameters? (The XML file and calculation results are in the attachment)

 

Regards,

Xiangjun Shan

 

从 Windows 版邮件发送

 

发件人: Stanisław Gepner <stanislaw.gepner@pw.edu.pl>
发送时间: Friday, February 18, 2022 10:12:06 PM
收件人: shan xiangjun <x_j_shan@outlook.com>
抄送: He, Wei <Wei.He@liverpool.ac.uk>; Nektar-users@imperial.ac.uk <nektar-users@imperial.ac.uk>
主题: Re: 回复: [Nektar-users] 2D direct stability analysis
 
Hi Shan,
Sorry for misleading you. I confused Direct Evolution Operator with the Coupled NS solver.

With the time stepping method, that you use, I think you might see a difference, depending on the time step and the number of steps taken, as this controls the period of a single iteration. I am not sure here, but I suspect that increasing this time, should give you more consistent results. If your problem is not very big you could try using the coupled solver with Arpack, and calculate more EV. The problem is it only works in serial.

Cheers,
Stan

18 lut 2022 12:19 shan xiangjun <x_j_shan@outlook.com> napisał(a):

Dear Stan and Wei,

 

Thanks a lot for your reply!

 

The XML files for base flow and stability analysis are attached.

 

I used the steady-state solver to calculate the base flow, and the resulting flow field is in the laminar flow state (the resulting file of base flow is also attached). Is it possible to ensure the convergence of the base flow in this way?

 

Regards,

 

Xiangjun Shan

 

 

从 Windows 版邮件发送

 

发件人: He, Wei <Wei.He@liverpool.ac.uk>
发送时间: Friday, February 18, 2022 6:23:35 PM
收件人: shan xiangjun <x_j_shan@outlook.com>
抄送: Nektar-users@imperial.ac.uk <nektar-users@imperial.ac.uk>
主题: Re: [Nektar-users] 2D direct stability analysis
 
Hi Xiangjun,

From my experience of computing airfoil flow, the leading mode maybe different if your base flow is not well converged. If you compare the eigenspetrum, I believe you will find there is no big difference. I think you can try a higher oder calculations or different Krylov space.

Regards,

Wei

On 18 Feb 2022, at 08:54, shan xiangjun <x_j_shan@outlook.com> wrote:

This email from x_j_shan@outlook.com 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 to disable email stamping for this address. 
 
Hi Everyone, 
 
For 2D direct stability analysis of Nektar + +, when raising the NumSteps, I get different eigenvectors and I'm not sure which is correct (as shown in the figure).
 
TimeStep  = 0.004                                           TimeStep = 0.004
NumSteps = 4250                                            NumSteps = 8250
<74181780A2BD44EFB8F01C3554F38D2E.png>  <872686C9FAC945B3B3D059AE5B7BA221.png>
 
Why does NumSteps affect the results and how do I set its value?
 
Looking forward to your reply!
 
Best regards,
 
Xiangjun Shan
 
 
从 Windows 版邮件发送
 
_______________________________________________
Nektar-users mailing list
Nektar-users@imperial.ac.uk
https://mailman.ic.ac.uk/mailman/listinfo/nektar-users