Hi Fabian,

you can actually run in verbose mode ./executable -v file.xml to see additional parameters, such as the operator evaluation.
Regarding the most appropriate hp space, this depends on the case.

For the Nan, I would suggest that you try to increase the number of quadrature points while keeping the number of modes fixed. A second possibility is that you start the simulation non-impulsively but using a ramp in time for the farfield boundary conditions. 

Regards,
Gianmarco

Gianmarco Mengaldo

Department of Aeronautics and Mathematics

Imperial College London 

South Kensington Campus

SW7 2AZ, London, UK

***Currently at Caltech



From: "Selbach, Fabian" <fabian.selbach@student.uni-siegen.de>
Subject: [Nektar-users] Default operator evaluation in Nektar++ and error "NaN found during time integration"
Date: 10 October 2016 at 14:09:28 BST
To: "nektar-users@imperial.ac.uk" <nektar-users@imperial.ac.uk>

Dear Nektar++ -Community,

please can you tell me the default operator evaluation in Nektar++? In the papers "From h to p efficiently"  there are mentioned three kind of types: 

1. global matrix approach
2. local matrix approach
3. sum-factorization

I want to choose the best hp/ configuration for my 3D simulation and this -of course- depends on the operator evaluation strategy. I think actually Nektar++ uses the global matrix approach by default, but I'm not sure by 100%.

My second question depends on the error message:

"Fatal   : Level 0 assertion violation
NaN found during time integration."

While using this configuration:
=======================================================================
            EquationType: EulerCFE
            Session Name: 3Dmeshcube_h10
            Spatial Dim.: 3
      Max SEM Exp. Order: 3
          Num. Processes: 24
          Expansion Dim.: 3
          Riemann Solver: HLLC
          Advection Type: 
         Projection Type: Discontinuous Galerkin
          Diffusion Type: 
               Advection: explicit
           AdvectionType: WeakDG
               Diffusion: explicit
               Time Step: 0.01
            No. of Steps: 10000
     Checkpoints (steps): 1000
        Integration Type: RungeKutta2_ImprovedEuler
            Problem Type: General
=======================================================================
Initial Conditions:
  - Field rho: rhoInf + (0.01 * exp(-0.077*(x^2 + y^2 + z^2)))
  - Field rhou: rhoInf * uInf
  - Field rhov: rhoInf * vInf
  - Field rhow: rhoInf * wInf
  - Field E: (pInf / (Gamma - 1)) + ((rhoInf/2)  * uInf^2) + ((0.01/(Gamma-1))*exp(-0.077*(x^2 + y^2 + z^2))) +( (0.01/2) * uInf^2 * exp(-0.077*(x^2 + y^2 + z^2)))

I already changed my time-step and the refinement of my mesh and i think this error message is because of oscillation of my solution. So maybe my mesh is not fine enough or i must increase p. But i want to use a fix h to p constellation to test the efficiency. Can anybody help me with this problem? I already tried to use the spectral hp dealiasing, the smooth advection and artificial viscosity, but it doesn't worked or helped...

Thanks for any answer.

Best regards
Fabian
_______________________________________________
Nektar-users mailing list
Nektar-users@imperial.ac.uk
https://mailman.ic.ac.uk/mailman/listinfo/nektar-users