May 2018 - Nek5000-users - mailman3.cels.anl.gov

dealias; mass matrix; interp_rstd
by nek5000-users＠lists.mcs.anl.gov 23 May '18

23 May '18

Dear Neks, I solve \partial A/\partial t = \nabla^2 A + u x Curl A, where A = (s1,s2,s3) I solve three scalar equations with respective components of u x Curl A as forcing terms in qvol. I would like to compute the dealiased product of u x Curl A I build my code based on kov_stokes_test.usr and subroutines in induct.f I compute local curl, project curl and U onto fine mesh, perform product and project the product back to coarse mesh using interp_rstd However, I am not sure about the multiplication of mass matrices for the product in these subroutines. I attach my usr, rea and SIZE files. Your suggestions are welcome. Thanks, Sandeep

1 1

Using Cartesian grid lookup table in Nekton
by nek5000-users＠lists.mcs.anl.gov 22 May '18

22 May '18

Hey Neks, I have a forcing function which is in the form of a look up table in X-Y cartesian grid. I would like to use this forcing function in the 2-D fluid momentum equation that Nekton solves. Is there a way to do this in the .usr file or elsewhere? Thanks, Saikat Saikat Mukherjee, PhD Student, Paul Research Group - http://www.me.vt.edu/mpaul/ Engineering Science and Mechanics, Virginia Tech.

1 0

Mesh optimisation for the cylinder case ext_cyl
by nek5000-users＠lists.mcs.anl.gov 22 May '18

22 May '18

Dear Nek5000 masters, I have been able to run the ext_cyl example. And, after postprocessing the case, it made me wonder if Nek5000 is able to run this case with a more optimized mesh? That is, refined just close to the cylinder and coarse away from it. Clearly, for this case, the refinement around the cylinder extends in sort of a “+” shape away from the cylinder. This refinement away from the cylinder is not necessary, specially towards the sides and towards the upstream. But even in the downstream the mesh could be (ideally) a bit coarser. Some other CFD codes are able to handle this problem by accepting refinement of the mesh just close to the cylinder (obviously at the surface of the refinement an element from the coarser side becomes neighbour of two or more elements on the refined side). I’m wondering if Nek5000 is already capable of handling this kind of refinement (or possibly any other solution/trick {maybe lowering the order of the elements themselves}) in order to avoid this typical problem of excessive refinement where it’s not needed/desired? Wondering and hoping, indeed, because this problem may become serious for some other configurations, for example for an array of cylinders, which then my cause the mesh to be refined all over the domain and not just in the “+” shape region. Thanks a lot, Alexis Dr Alexis Espinosa Supercomputing Specialist Pawsey Supercomputing Centre 26 Dick Perry Ave Kensington WA 6151 Australia Tel +61 xxxx xxxx Email alexis.espinosa(a)pawsey.org.au Web www.pawsey.org.au From: Alexis Espinosa Gayosso <alexis.espinosa(a)pawsey.org.au> Date: Tuesday, 8 May 2018 at 4:46 pm To: <nek5000-users(a)lists.mcs.anl.gov> Subject: Compiling Problems in "Running your very first simulation" Hello, The title of my thread is the same as for an already existing thread (26-Sept-2017) because: I receive very similar warning messages when compiling nek5000 (./maketools all). Lots of warnings! >From your answer on that thread it seems that you do not worry at all for the warning messages. Should we really not care about them? How to know which warnings are not of concern and which other are? Can you post your own .log file to see which warning messages are considered normal for you? Or how do you check everything went fine for the compilation? Anyway, I will start running the examples in order to learn about the usage of the code. And I guess any abnormality would appear there. Many thanks, Alexis Dr Alexis Espinosa Supercomputing Specialist Pawsey Supercomputing Centre 26 Dick Perry Ave Kensington WA 6151 Australia Tel +61 xxxx xxxx Email alexis.espinosa(a)pawsey.org.au Web www.pawsey.org.au

1 1

Re: [Nek5000-users] ffx ffy ffz from current run
by nek5000-users＠lists.mcs.anl.gov 21 May '18

21 May '18

Hi Paul, The shear Reynolds number of my case is 361 which corresponding to 11700. Can you please have a look my attached usr file? https://drive.google.com/open?id=1hFLyRNQ55vXGcT9uyMNvVNt7kX6dBRXE Kind regards, Zhai

1 0

Re: [Nek5000-users] ffx ffy ffz from current run
by nek5000-users＠lists.mcs.anl.gov 21 May '18

21 May '18

Dear Paul, After last request email ( https://lists.mcs.anl.gov/mailman/htdig/nek5000-users/2018-March/005178.html) the ffz value of 0.01525 was obtained by 'grep volf logfile'. Then I used this value as ffz in usr file with changing para54 and 55 as 0 in the same time. The case was started from previous stop which is based on keeping flowrate. After long time running (delt t*=200), the streamwise velocity was decreased compared with previous running . May I know what I can do to get the turbulence and velocity back? Kind regards, Zhai

1 1

mesh import options
by nek5000-users＠lists.mcs.anl.gov 21 May '18

21 May '18

Hi all, Is exo2nek still the best way to import meshes with relatively complex geometries to nek5000? If yes, can someone please help me out regarding exo2nek import: 1. I made my mesh using Pointwise and exported to exodusII format, but exo2nek throws the following error: """ EXODUS: ERROR: Attempting to open the netcdf-4 file: 'pipe.exo' . Either the netcdf library does not support netcdf-4 or there is a filesystem or some other issue ERROR: cannot open file pipe.exo """ 2. I exported exodus from ICEM-CFD, here the error from exo2nek reads: """ element block id = 1 element type = HEX num_elem_in_block = 24500 num_nodes_per_elem = 8 ERROR: Only HEX27 elements are allowed in a 3D mesh! num_nodes_per_elem= 8 Note: The following floating-point exceptions are signalling: IEEE_DENORMAL """ 3. The popular option looks exporting exodus file from Gambit with hex27 elements, however I was not able to find Gambit meshing software. Is it open source now or comes in with Fluent? I've Fluent 18.0 license at my university. *Cheers,Vishal*

1 1

Re: [Nek5000-users] Problem with Visit in-situ
by nek5000-users＠lists.mcs.anl.gov 18 May '18

18 May '18

Dear all, I downloaded the newest version of Nek5000 and the code compiles with Visit. I noticed the old version was missing the definition of the "visitgetmixedvariable" in the visit.f file. However, the code is still not working correctly. I'm using the eddy_uv as example. I am able to connect to the simulation using VisIt but when I try to add a pseudocolor plot of "x_velocity", VisIt throws the following error: Pseudocolor: () viewer: The pipeline object is being used improperly: The simulation returned a handle for an object other than a variable. The mesh is plotted correctly, but I cannot visualize anything else. Any ideas? Thanks, Juan Diego

1 0

Re: [Nek5000-users] Curved boundary
by nek5000-users＠lists.mcs.anl.gov 18 May '18

18 May '18

Note, all triple digit versions like 17.0.4 and 17.0.0 are _internal_ development versions. The only official release up to now is v17.0 (available as tarball only). Anyway, yes even in v17.0 we disabled certain features for PnPn as some combinations are not yet supported or well tested. It's very likely that some of the might work out of the box but we offically we don't support them (at least for now). Stefan. -----Original message----- > From:[email protected] <nek5000-users(a)lists.mcs.anl.gov> > Sent: Friday 18th May 2018 13:58 > To: nek5000-users(a)lists.mcs.anl.gov > Subject: [Nek5000-users] Curved boundary > > Hi all, > > I have a case with a curved boundary with SYM boundary condition, stress > formulation and PN/PN. > > In the file core/bdry.f there is a function called CHKCBC, which checks > the consistency of boundary conditions. > > In the latest release (mver=17.0.4), this routine requires stress > formulation AND PN/PN-2 for a non-aligned boundary, whereas in > mver=17.0.0 stress formulation with PN/PN was accepted. > > My question is whether this change was intentional? > > The test > "IF ( .NOT.IFSTRS ) CALL CHKCBC" > in function SETLOG in v17.0.0 was replaced by > "if (ifstrs .and. .not.ifsplit) return" > inside CHKCBC in v17.0.4. > > As a test, I commented out the check in CHKCBC, and the simulation seems > to run just fine with stress formulation and PN/PN. > > Thanks, > Carl Erik, > FFI, Norway > _______________________________________________ > Nek5000-users mailing list > Nek5000-users(a)lists.mcs.anl.gov > https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users >

1 2

Curved boundary
by nek5000-users＠lists.mcs.anl.gov 18 May '18

18 May '18

Hi all, I have a case with a curved boundary with SYM boundary condition, stress formulation and PN/PN. In the file core/bdry.f there is a function called CHKCBC, which checks the consistency of boundary conditions. In the latest release (mver=17.0.4), this routine requires stress formulation AND PN/PN-2 for a non-aligned boundary, whereas in mver=17.0.0 stress formulation with PN/PN was accepted. My question is whether this change was intentional? The test "IF ( .NOT.IFSTRS ) CALL CHKCBC" in function SETLOG in v17.0.0 was replaced by "if (ifstrs .and. .not.ifsplit) return" inside CHKCBC in v17.0.4. As a test, I commented out the check in CHKCBC, and the simulation seems to run just fine with stress formulation and PN/PN. Thanks, Carl Erik, FFI, Norway

1 0

Temperature Perturbation Zone
by nek5000-users＠lists.mcs.anl.gov 17 May '18

17 May '18

Hi, I'm setting a 3D box to perform natural convection problems in a closed cavity. After trying several method i can't manage to define a zone with a variable temperature perturbation in the usrchk() routine. Let's say, that temp = temp + A*sin(w*time) in a square of boundaries [x_min, x_max] x [y_min, y_max] x [z_min, z_max] for x, y, z. Any advice for implementing this? Regards, *Nicolás Andrés Órdenes Leiva* Estudiante de Ingeniería Civil Mecánica Universidad Técnica Federico Santa María Tel: +56 9 7619 2857

1 4