October 2017 - Nek5000-users - mailman3.cels.anl.gov

wavelet decomposition
by nek5000-users＠lists.mcs.anl.gov 31 Oct '17

31 Oct '17

Dear all, I would like to perform a wavelet decomposition on my DNS data of stratified turbulence generated using Nek5000. The details of the decomposition to extract coherent vorticies is explained in this paper: Farge, M., Schneider, K., Pellegrino, G., Wray, A. A., & Rogallo, R. S. (2003). Coherent vortex extraction in three-dimensional homogeneous turbulence: Comparison between CVS-wavelet and POD-Fourier decompositions. *Physics of Fluids*, *15*(10), 2886-2896. I am wondering if anyone in our community has had any experience with integrating the wavelet analysis with Nek5000 in the post-processing mode? If not, would you suggest performing such a decomposition in the VisIt Python Interface using a python package like PyWavelets for example? Thanks a lot for the help, Regards, Hesam

1 1

Re: [Nek5000-users] par file and moving mesh
by nek5000-users＠lists.mcs.anl.gov 27 Oct '17

27 Oct '17

One more update on this matter. I generated a mesh with cubit that I exported in a exodus format and in a msh format. I then did the following: - in the first case, I used the exo2nek script to generate the re2 file, and run Nek5000 to get a rea file by calling ‘gen_rea’. - I used the msh2nek python script to generate a rea file. My geometry has a moving boundary conditions and the mesh motion is solved by Nek5000 in my current setting. I observe the following when running all above rea files with the same Nek5000 compiled usr file: - when running the Nek5000 session with the re2 file generated from the exo2nek script, the mesh is static when visualized with Visit. The log file shows that Nek5000 is solving for the mesh motion, though. - when running the Nek5000 session with the rea file generated with the gen_rea call, the mesh moves when visualized with Visit, but Nek5000 exits the run and writes a file named ‘xyz***’. - when running the Nek5000 session with the rea file generated with the python script, the mesh moves, the solution looks good and Nek5000 runs all the way to the final time specified in the rea file. When running the Nek5000 session with the re2 file generated with the exo2nek script, I set the boundary condition in the usrdat subroutine. At first glance, it seems to be the right approach since the boundary conditions are correctly labeled in the rea file that is generated with gen_rea. But because Nek5000 exits with an error message when I run the rea file coming from the gen_rea call, I am not so sure. Should I be setting the boundary conditions at a different initialization stage? Note that I used the exo2nek script for other geometries that have more common boundary conditions (v and O), and I did not run into any issue. This problem seems to be specific to the moving mesh and boundary conditions capabilities, I think. Once again, I would appreciate any help on this matter. Thanks, Marco On Oct 24, 2017, at 12:55 PM, Delchini, Marc-Olivier G. <delchinimg(a)ornl.gov<mailto:[email protected]>> wrote: I specify the boundary conditions in the usr file using the following piece of code: do i=1,nelt do j=1,2*ndim if(bc(5,j,i,1).eq.1) then cbc(j,i,1)='mv ' elseif(bc(5,j,i,1).eq.2) then cbc(j,i,1)='W ' elseif(bc(5,j,i,1).eq.3) then cbc(j,i,1)='W ' elseif(bc(5,j,i,1).eq.4) then cbc(j,i,1)='W ' else cbc(j,i,1)='E ' endif enddo enddo Marco On Oct 24, 2017, at 12:39 PM, nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> wrote: How do you specify the BC in the exo case? On 24 Oct 2017, at 18:05, "nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]>" <nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]>> wrote: Stefan, thank for your reply. My mesh moves if I write the mesh in a rea file. If I use the mesh generated from exo2nek script, the mesh does not move. Marco > On Oct 24, 2017, at 11:59 AM, nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> wrote: > > Marco, > > please note that the .par file is still experimental and not as robust as the good old .rea. > It looks like I forgot to support the viscosity and tolerance for the mesh elasticity solver - so the default values 0.4 and 0.02 will be used. Anyway, your mesh should be moving unless there is something wrong with the BC in the exo case. > > Cheers, > Stefan > > -----Original message----- >> From:[email protected]<mailto:[email protected]> <nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]>> >> Sent: Monday 23rd October 2017 22:15 >> To: nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> >> Subject: Re: [Nek5000-users] par file and moving mesh >> >> Hi all, >> >> I have an update on this issue. >> >> The mesh I am using was generated in Cubit and then converted to the Nek5000 re2 format using the exo2nek script. I call the function ‘gen_rea’ to make Nek5000 write the mesh in a rea file. My mesh has a moving boundary condition, and the mesh motion is computed by the ALE solver. I then do the following: >> 1) I run my session using the mesh read from the re2 file that comes from the exo2nek script. >> 2) I run my session using the rea file that was generated by the function call ‘gen_rea'. >> >> In the first case, the output file does not show the mesh motion even though the velocity profile seems correct. In the second case, the mesh motion shows in the output file when I visualize the solution with Visit. >> >> Is this a bug, or simply a misuse of the exo2nek script? >> >> Marco >> >>> On Oct 19, 2017, at 8:51 AM, nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> <Nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]>> wrote: >>> >>> Hi all, >>> >>> I am trying to switch from a rea file to a par file and I am having difficulties with the moving mesh capabilities. When visualizing the numerical solution, the mesh is not moving, whereas it was moving with a previous version of Nek5000. >>> >>> I copied the par file from the my_cyl examples as a start point. The main different between my session and the mv_cyl session is the way the mesh motion is handled. I believe the mesh velocity is provided by the user through a function, whereas in my session the mesh velocity is solved for. >>> >>> Below is the par file I am currently using: >>> >>> # >>> # nek parameter file >>> # >>> [GENERAL] >>> #startFrom = restart.fld >>> stopAt = numSteps >>> endTime = 0 >>> numSteps = 1000 >>> >>> dt = 1e-03 >>> timeStepper = bdf #bdf #char #steady >>> tOrder = 3 >>> variableDt = no >>> maxCFL = 2.0 >>> >>> writeControl = timeStep #runTime >>> writeInterval = 50 >>> >>> dealiasing = yes >>> userParam01 = 1.16 >>> userParam02 = 1.8550800E-05 >>> >>> [PROBLEMTYPE] >>> lowMachNumber = no >>> variableProperties = yes >>> dp0dt = no >>> >>> [MESH] >>> motion = yes >>> meshVelocity = no >>> >>> [PRESSURE] >>> residualTol = 1e-10 >>> residualProj = no >>> #preconditioner = amg >>> >>> [VELOCITY] >>> residualTol = 1e-12 >>> residualProj = no >>> >>> >>> Any help will be appreciated. >>> >>> Thanks, >>> >>> Marco >>> >>> _______________________________________________ >>> Nek5000-users mailing list >>> Nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> >>> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users >>> >> >> _______________________________________________ >> Nek5000-users mailing list >> Nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> >> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users >> > _______________________________________________ > Nek5000-users mailing list > Nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> > https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users _______________________________________________ Nek5000-users mailing list Nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users _______________________________________________ Nek5000-users mailing list Nek5000-users(a)lists.mcs.anl.gov<mailto:[email protected]> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users

1 4

Re: [Nek5000-users] amg solver
by nek5000-users＠lists.mcs.anl.gov 27 Oct '17

27 Oct '17

Thanks a lot, Stefan, for your reply. But I use the amg_matlab2/run script instead of amg_hypre, and it seems to produce the same files: amg.dat, amg_Af*, amg_W.dat What is the difference between these approaches? Regards, Vatslav.

1 0

Re: [Nek5000-users] amg solver
by nek5000-users＠lists.mcs.anl.gov 27 Oct '17

27 Oct '17

Note, the AMG section of the documentation is outdated (see issues on NekDoc). To use the AMG solver set p40=1 (for .rea) or solver=semg_amg (in pressure section for .par file). This will dump all required AMG setup files. Then run amg_hypre (part of our tools) in your case directory. Now just run Nek5000 again. Cheers, Stefan -----Original message----- > From:[email protected] <nek5000-users(a)lists.mcs.anl.gov> > Sent: Friday 27th October 2017 12:19 > To: nek5000-users(a)lists.mcs.anl.gov > Subject: [Nek5000-users] amg solver > > Dear Neks, > > Recently I decide to use the amg solver. > I had done all preparations from github documentation, but it did not work for the first time. > After this I look throw the forum and catch up the topic about amg dump. > It was said that p40 in rea file is bound with amg solver, although official manual says that p40 "NOT IN USE". > Since I set the p40 to 1 the amg solver starts to work (amg dump and after amg solver). > So I want to clarify, what exactly does this p40 parameter ? > And perhaps is there any additional flag or parameters (except ifamg and ifamg_dump) that I need to do for make amg solver works correctly in my simulation? > <br clear="all" />All Regards, Vatslav. > _______________________________________________ > Nek5000-users mailing list > Nek5000-users(a)lists.mcs.anl.gov > https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users

1 0

amg solver
by nek5000-users＠lists.mcs.anl.gov 27 Oct '17

27 Oct '17

Dear Neks, Recently I decide to use the amg solver. I had done all preparations from github documentation, but it did not work for the first time. After this I look throw the forum and catch up the topic about amg dump. It was said that p40 in rea file is bound with amg solver, although official manual says that p40 "NOT IN USE". Since I set the p40 to 1 the amg solver starts to work (amg dump and after amg solver). So I want to clarify, what exactly does this p40 parameter ? And perhaps is there any additional flag or parameters (except ifamg and ifamg_dump) that I need to do for make amg solver works correctly in my simulation? All Regards, Vatslav.

1 0

Reynolds for pipe
by nek5000-users＠lists.mcs.anl.gov 26 Oct '17

26 Oct '17

Hi, May I know for shear Reynolds number (Re_t) equals to 361, whether setting p02 VISCOS in rea file to be -11700 and the pipe radius equals to 1 in the mesh part is correct. Kind regards, ZJ

1 2

variable viscosity formulation
by nek5000-users＠lists.mcs.anl.gov 25 Oct '17

25 Oct '17

Hi Nek Forum, I am using a variable viscosity formulation for a Rayleigh-Benard problem. I get the following error after some time : ERROR: Non-positive diffusivity ( -0.481 ) specified for field 1, group 0 element Please let me know what might be wrong. Thank you Swarandeep

1 1

Nek5000 Hackathon, November 12-14, 2017
by nek5000-users＠lists.mcs.anl.gov 25 Oct '17

25 Oct '17

Dear Nek Developers and Users, We are excited to announce the first-ever Nek5000 Hackathon. When: Nov. 12 (1 PM, Sunday) -- Nov. 14 (5pm, Tuesday), 2017 Where: NCSA Building, University of Illinois, Urbana-Champaign (UIUC), IL, USA Hackathon Goal: The event will bring together users and developers with the intent to help new and current users bring their applications into production mode. It will also give users a chance to provide the development team with feedback on what features or enhancements they would like to see in future releases. Format: We are asking each of the participants (or as a team) to develop a small example demonstrating the phenomena of interest that could be shared with the rest of the user community through the NekExamples Suite. There will not be a formal presentation agenda. We anticipate, however, several rounds of short presentations describing the problems of interest, the features that make the problem interesting and challenging, and the anticipated ways in which the Nek team can help. Please bring your laptops and come prepared to access your usual work sites remotely, from UIUC. Cost: There is no formal registration fee. Lunches will be covered. We may collect a nominal fee for group dinners on Sunday and Monday. Registration: In order to anticipate the number of attendees, we will post a registration site in the coming week. Please stay tuned for details. Hotels: There are several hotels in the area that are a short walk from the workshop site (Siebel Center, Day 1; NCSA Days 2 & 3). Hampton Inn on University ~ 5 minute walk. Illini Union ~ 10 minute walk. Towne Place Suites by Marriott, Champaign 15 minutes I Hotel - 2 miles Sylvia's Irish Inn B&B 10 minutes We look forward to seeing you next month! Sincerely, The Nek Development Team

1 1

Restarting with Periodic boundary Condition
by nek5000-users＠lists.mcs.anl.gov 25 Oct '17

25 Oct '17

Hello Neks, I have a small (4x4x1) cubic domain, which is periodic in Y. No slip boundaries in X and Z. This is to generate a Rayleigh Benard convection flowfield with Z being the convective direction. The flow is driven by hot sidewalls in X. My problem is when I restart the flowfiled from a particular flowfiled file using the same .rea, .usr files and everything, the periodicity in the Y direction seems to be gone. My originial flowfileds are periodic in Y and have the same value of Velocity Magnitude through all the Y slices ( as can be seen with VisIt). But my restarted files have different values of velocity magnitude in different slices of Y, which should not be the case. This seems to suggest something going wrong during restarting for the periodic bc side. I haven't noticed any change in the Velocity if I have no-slip in Y instead of Periodic. I would appreciate any input on this problem. Thanks, Saikat Saikat Mukherjee, PhD Student, Paul Research Group - http://www.me.vt.edu/mpaul/ Engineering Science and Mechanics, Virginia Tech.

1 1

Re: [Nek5000-users] par file and moving mesh
by nek5000-users＠lists.mcs.anl.gov 24 Oct '17

24 Oct '17

Marco, please note that the .par file is still experimental and not as robust as the good old .rea. It looks like I forgot to support the viscosity and tolerance for the mesh elasticity solver - so the default values 0.4 and 0.02 will be used. Anyway, your mesh should be moving unless there is something wrong with the BC in the exo case. Cheers, Stefan -----Original message----- > From:[email protected] <nek5000-users(a)lists.mcs.anl.gov> > Sent: Monday 23rd October 2017 22:15 > To: nek5000-users(a)lists.mcs.anl.gov > Subject: Re: [Nek5000-users] par file and moving mesh > > Hi all, > > I have an update on this issue. > > The mesh I am using was generated in Cubit and then converted to the Nek5000 re2 format using the exo2nek script. I call the function ‘gen_rea’ to make Nek5000 write the mesh in a rea file. My mesh has a moving boundary condition, and the mesh motion is computed by the ALE solver. I then do the following: > 1) I run my session using the mesh read from the re2 file that comes from the exo2nek script. > 2) I run my session using the rea file that was generated by the function call ‘gen_rea'. > > In the first case, the output file does not show the mesh motion even though the velocity profile seems correct. In the second case, the mesh motion shows in the output file when I visualize the solution with Visit. > > Is this a bug, or simply a misuse of the exo2nek script? > > Marco > > > On Oct 19, 2017, at 8:51 AM, nek5000-users(a)lists.mcs.anl.gov <Nek5000-users(a)lists.mcs.anl.gov> wrote: > > > > Hi all, > > > > I am trying to switch from a rea file to a par file and I am having difficulties with the moving mesh capabilities. When visualizing the numerical solution, the mesh is not moving, whereas it was moving with a previous version of Nek5000. > > > > I copied the par file from the my_cyl examples as a start point. The main different between my session and the mv_cyl session is the way the mesh motion is handled. I believe the mesh velocity is provided by the user through a function, whereas in my session the mesh velocity is solved for. > > > > Below is the par file I am currently using: > > > > # > > # nek parameter file > > # > > [GENERAL] > > #startFrom = restart.fld > > stopAt = numSteps > > endTime = 0 > > numSteps = 1000 > > > > dt = 1e-03 > > timeStepper = bdf #bdf #char #steady > > tOrder = 3 > > variableDt = no > > maxCFL = 2.0 > > > > writeControl = timeStep #runTime > > writeInterval = 50 > > > > dealiasing = yes > > userParam01 = 1.16 > > userParam02 = 1.8550800E-05 > > > > [PROBLEMTYPE] > > lowMachNumber = no > > variableProperties = yes > > dp0dt = no > > > > [MESH] > > motion = yes > > meshVelocity = no > > > > [PRESSURE] > > residualTol = 1e-10 > > residualProj = no > > #preconditioner = amg > > > > [VELOCITY] > > residualTol = 1e-12 > > residualProj = no > > > > > > Any help will be appreciated. > > > > Thanks, > > > > Marco > > > > _______________________________________________ > > Nek5000-users mailing list > > Nek5000-users(a)lists.mcs.anl.gov > > https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users > > > > _______________________________________________ > Nek5000-users mailing list > Nek5000-users(a)lists.mcs.anl.gov > https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users >

1 3