April 2019 - Nektar-users - imperial.ac.uk

Element with max CFL
by Vishal Saini 02 Apr '20

02 Apr '20

Hi all, Quick question: Has anyone tried to check the location of a particular element in the mesh? A log file reports the maximum (?) CFL and the corresponding element as: "CFL: 0.0337915 (in elmt 94488)"? Is there a quick way to locate "elmt 94488"? Cheers, Vishal

2 2

blood flow simulation
by zxiong＠sas.upenn.edu 10 May '19

10 May '19

Dear all, I am trying to do a cardiovascular simulation by using Incompressible N-S solver in Nektar++. Here I have two questions about setting boundary conditions. First, how can I set RCR(windkessel model) boundary condition if I use Incompressible N-S solver? Is there any similar input file I can find? Second, do I need to make all the variable dimensionless when I set boundary conditions? Thank you very much! Best, Zhen

3 3

problem with FieldConvert interpfield
by A.P. 22 Apr '19

22 Apr '19

Hello, I try to convert a Z-homogenious fld file to an another mesh by FieldConvert module 'interpfield' and recieve this error message: Fatal : Level 0 assertion violation Failed to type cast m_exp[0] terminate called after throwing an instance of 'ErrorUtil::NekError' what(): Level 0 assertion violation Failed to type cast m_exp[0] 2D fields are converted OK. What can be done to overcome the error? Alexander.

2 2

Hi all (( problem in prob points in 3D ))
by Moto Light 21 Apr '19

21 Apr '19

Hi All, I am setting prob line on 2.5D for 4D spanwise, for using 32 modes or 64 modes it is ok, but when i am using 48 modes it changes the points location like this: *Reseting History point from z = 2.83333 to z = 2.75* *Reseting History point from z = 2.91667 to z = 2.91667* *Reseting History point from z = 3.08333 to z = 3* *Reseting History point from z = 3.16667 to z = 3.16667* *Reseting History point from z = 3.33333 to z = 3.25* *Reseting History point from z = 3.41667 to z = 3.41667* *Reseting History point from z = 3.58333 to z = 3.5* *Reseting History point from z = 3.66667 to z = 3.66667* *Reseting History point from z = 3.83333 to z = 3.75* *Reseting History point from z = 3.91667 to z = 3.91667* *any suggestions?* *Regards,* *Reda*

2 1

compressible flow solver scalability with fortran subroutines
by Arash Nouri 20 Apr '19

20 Apr '19

Hi all, I've added an equation of state and a chemical reaction to the compressible flow solver via couple of FORTRAN subroutines. In my 2D simulations, everything is cool and I've almost linear scalability up to 2K processors. However, my 3D simulations are surprisingly slow. Do you have any idea about probable reasons? What I basically do for the chemical reaction is: I call a FORTRAN subroutine in DoOdeRhs to get the required forcing terms for energy and species equations. Best, Arash --------------------------------- Arash G. Nouri Research Scientist Computational Modeling & Simulation (http://cmsp.pitt.edu/) 1127 Benedum Hall University of Pittsburgh Pittsburgh, PA 15261, USA E-mail: arn36(a)pitt.edu

2 1

blood flow simulation
by zxiong＠sas.upenn.edu 18 Apr '19

18 Apr '19

Dear all, I am trying to do a cardiovascular simulation by using Nektar++. Here I have two questions about setting boundary conditions. First, how can I set RCR(windkessel model) boundary condition in nektar++? Is there any similar input file I can find? Second, do I need to make all the variable dimensionless when I set boundary conditions? Thank you very much! Best, Zhen

1 0

Add Smagorinsky model to IncNSSolver
by Vishal Saini 18 Apr '19

18 Apr '19

Hi all, Please consider me as a beginner in editing the Nektar++ solvers. I'd like to implement Sub-Grid-Scale Smagorinsky model into IncNavierStokesSolver. This would require: 1. local element size, 2. calculation of the velocity gradient at each point in physical space, 3. adding turbulent viscosity nu_turb=func( size_element, grad(v) ) to the viscosity in physical space. I had a look into the code focusing on my requirements and have a few questions: 1. Re documentation, is there any developer guide for the IncNavierStokesSolver as a starting point? 2. Re the implementation, I'm using the velocity correction scheme. Please suggest if the strategy below makes sense (performance-wise): 2.a) For the solution points separation in space, repeat the same as in SVVVarDiffCoeff() function to calculate h/p: { h = max(h, exp3D->GetGeom3D()->GetEdge(i)->GetVertex(0)->dist( *(exp3D->GetGeom3D()->GetEdge(i)->GetVertex(1)))); } for(int i = 0; i < 3; ++i) { p = max(p,exp3D->GetBasisNumModes(i)-1); } 2.b) Obtain grad(v) using PhysDeriv(space, var, gradvar). However, I need some help here. Is it possible to access different components of the gradients (du_dx, du_dy, du_dz etc.) of the advected fields without going into the loops as done in ProcessModules/ProcessGrad.cpp? 2.c) Add the above changes in VelocityCorrectionScheme.cpp as commented: /** * Solve velocity system */ void VelocityCorrectionScheme::v_SolveViscous( const Array<OneD, const Array<OneD, NekDouble> > &Forcing, Array<OneD, Array<OneD, NekDouble> > &outarray, const NekDouble aii_Dt) { StdRegions::ConstFactorMap factors; StdRegions::VarCoeffMap varCoeffMap = StdRegions::NullVarCoeffMap; MultiRegions::VarFactorsMap varFactorsMap = MultiRegions::NullVarFactorsMap; AppendSVVFactors(factors,varFactorsMap); // // // Smagorinsky model Start // // // // Find ele_size i.e. h/p as in SVVVarDiffCoeff() // Find du_dxi, dv_dxi, dw_dxi, dAnotherField_dxi, where i=1,2,3 // nu_t = sq(const*ele_size)*sqrt(2.0*S_ijS_ij) // add nu_t field to m_diffCoeff // // // Smagorinsky model End // // // // Solve Helmholtz system and put in Physical space for(int i = 0; i < m_nConvectiveFields; ++i) { // Setup coefficients for equation factors[StdRegions::eFactorLambda] = 1.0/aii_Dt/m_diffCoeff[i]; m_fields[i]->HelmSolve(Forcing[i], m_fields[i]->UpdateCoeffs(), NullFlagList, factors, varCoeffMap, varFactorsMap); m_fields[i]->BwdTrans(m_fields[i]->GetCoeffs(),outarray[i]); } } 3. Since the velocity field is not C1 continuous at element boundaries, does anyone see a problem due to that? General comments are welcome on the methodology. Sorry for a long email. BR, Vishal --- *Vishal SAINI* PhD Student Loughborough University

1 0