Re: [Nektar-users] stabilisation/dealiasing in Compressible solver
Hi Spencer and other members, Thanks Spencer for your reply on the issue. I think the problem was under-resolution of the vortex. Projection error seems to be relatively small. Also, the compressible flow tutorial was helpful to set the boundary conditions properly. I have another (very basic) query regarding determining the CFL (u*delT/delX) for the case. I suppose that for N=2, delX is same as the element size, but I'm not sure on how to define delX in case of N>2. Should I take into account the point distribution inside the element (GLL, Jacobi etc.) and then delX=min(dist. between the points), or should I consider a uniform point distribution inside the element (delX= element_size/(N-1))? I'm using compressible solver (DG and Runge-Kutta 4-step time integration). P.S. I do not understand the fact that if there is a non-uniform distribution of quadrature points inside an element why do I see a uniform distribution in the post-processing softwares (paraview, TecPlot etc.). Any input is highly appreciated. Sincerely, Vishal On Thu, Sep 21, 2017 at 10:13 PM, Spencer Sherwin <ssherwin69@googlemail.com
wrote:
Hi Vishal,
Thank for your interest in Nektar++.
I have to confess I am not quite sure what the status of SpectralHpDealiasing is in the compressible solver (@Douglas, Dani, Gian: can you clarify). I do believe we may have investigated it for some of the convection operators but do not recall which ones.
SVV is definitely only setup for the incompressible flow solver since the DG naturally has this type of property.
One can however solve with a higher number of quadrature points and a more brute force way of realising. It would be interesting to know if this sorts out the issue you are observing. It might also be that the initial projection is leading to the poor approximation of the initial conditions and then being advected over the domain. (perhaps see Expansions in the user-guide for this).
We have been tying to put together a tutorial on the compressible flow solver so that we can help also demonstrate the use of artificial viscosity which can also be very useful in starting up flows. I hope to get this fully online some time soon!
Cheers, Spencer.
PS The MSc student who wrote the tutorial has also been doing some flow studies which might be of interest. If so I can let you have a copy of her report.
On 18 Sep 2017, at 17:55, Vishal Saini <vishal.saini.nitj@gmail.com> wrote:
Dear all,
I've just started using high-order CFD (using Nektar++). To begin with, I've just been running the Isentropic vortex advection case in a 50x20 domain using the IsentropicVortex test case provided. The issue is that I've continually been observing spurious oscillations (a figure is attached) in velocity field. Increasing the resolution improves the situation a little, but I could not completely get rid of the oscillations.
Next, I wanted to use dealiasing. However, the simulations do not even start when I use SVV and SPECTRALHPDEALIASING. Plus, I could not find any stabilization technique implementation in "nektar++-4.4.0/solvers/ CompressibleFlowSolver/Utilities" folder (which is the case in Incompressible solver). Can anyone clarify if any dealiasing technique is implemented in Compressible Solver and if there are any future plans on that?
Cheers and regards, Vishal
--- *Vishal SAINI* Master of Research, University of Cambridge. Master in Turbulence EC Lille, ENSIP and ENSMA France. Contact: vs434@cam.ac.uk, vishal.saini.nitj@gmail.com <vishal.saini.nitj@gmail.com> (+44) 7 459 186 139 (UK) (+33) 7 58 24 84 02 <+33%207%2058%2024%2084%2002> (France) <Screenshot from 2017-09-18 17-44-57.png>_________________ ______________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
Spencer Sherwin FREng, FRAeS McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ
s.sherwin@imperial.ac.uk +44 (0) 20 759 45052
Hi Vishal, Hi Spencer and other members, Thanks Spencer for your reply on the issue. I think the problem was under-resolution of the vortex. Projection error seems to be relatively small. Also, the compressible flow tutorial was helpful to set the boundary conditions properly. I have another (very basic) query regarding determining the CFL (u*delT/delX) for the case. I suppose that for N=2, delX is same as the element size, but I'm not sure on how to define delX in case of N>2. Should I take into account the point distribution inside the element (GLL, Jacobi etc.) and then delX=min(dist. between the points), or should I consider a uniform point distribution inside the element (delX= element_size/(N-1))? I'm using compressible solver (DG and Runge-Kutta 4-step time integration). This is a slightly tricky question. The GLL type point is snore consistent since the time step tends to scale as h/N^2 and since the GLL points cluster at a rate which is of order (O(N^2)) toward the end this gives a semi-reasonable estimate. However for lower P orders often this is a bit pessimistic and so using h/N is OK. P.S. I do not understand the fact that if there is a non-uniform distribution of quadrature points inside an element why do I see a uniform distribution in the post-processing softwares (paraview, TecPlot etc.). When we post-process using FieldConvert to paraview or Tecplot we interpolate to an equispaced mesh which is why you observe it in this format. Cheers, Spencer. On Thu, Sep 21, 2017 at 10:13 PM, Spencer Sherwin <ssherwin69@googlemail.com<mailto:ssherwin69@googlemail.com>> wrote: Hi Vishal, Thank for your interest in Nektar++. I have to confess I am not quite sure what the status of SpectralHpDealiasing is in the compressible solver (@Douglas, Dani, Gian: can you clarify). I do believe we may have investigated it for some of the convection operators but do not recall which ones. SVV is definitely only setup for the incompressible flow solver since the DG naturally has this type of property. One can however solve with a higher number of quadrature points and a more brute force way of realising. It would be interesting to know if this sorts out the issue you are observing. It might also be that the initial projection is leading to the poor approximation of the initial conditions and then being advected over the domain. (perhaps see Expansions in the user-guide for this). We have been tying to put together a tutorial on the compressible flow solver so that we can help also demonstrate the use of artificial viscosity which can also be very useful in starting up flows. I hope to get this fully online some time soon! Cheers, Spencer. PS The MSc student who wrote the tutorial has also been doing some flow studies which might be of interest. If so I can let you have a copy of her report. On 18 Sep 2017, at 17:55, Vishal Saini <vishal.saini.nitj@gmail.com<mailto:vishal.saini.nitj@gmail.com>> wrote: Dear all, I've just started using high-order CFD (using Nektar++). To begin with, I've just been running the Isentropic vortex advection case in a 50x20 domain using the IsentropicVortex test case provided. The issue is that I've continually been observing spurious oscillations (a figure is attached) in velocity field. Increasing the resolution improves the situation a little, but I could not completely get rid of the oscillations. Next, I wanted to use dealiasing. However, the simulations do not even start when I use SVV and SPECTRALHPDEALIASING. Plus, I could not find any stabilization technique implementation in "nektar++-4.4.0/solvers/CompressibleFlowSolver/Utilities" folder (which is the case in Incompressible solver). Can anyone clarify if any dealiasing technique is implemented in Compressible Solver and if there are any future plans on that? Cheers and regards, Vishal --- Vishal SAINI Master of Research, University of Cambridge. Master in Turbulence EC Lille, ENSIP and ENSMA France. Contact: vs434@cam.ac.uk, vishal.saini.nitj@gmail.com<mailto:vishal.saini.nitj@gmail.com> (+44) 7 459 186 139 (UK) (+33) 7 58 24 84 02<tel:+33%207%2058%2024%2084%2002> (France) <Screenshot from 2017-09-18 17-44-57.png>_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk<mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users Spencer Sherwin FREng, FRAeS McLaren Racing/Royal Academy of Engineering Research Chair, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus London SW7 2AZ s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0) 20 759 45052 _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk<mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users Spencer Sherwin FREng, FRAeS Head, Aerodynamics, Professor of Computational Fluid Mechanics, Department of Aeronautics, Imperial College London South Kensington Campus, London, SW7 2AZ, UK s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> +44 (0)20 7594 5052 http://www.imperial.ac.uk/people/s.sherwin/
participants (2)
-
Sherwin, Spencer J
-
Vishal Saini