AL: some manuals

Author: andrealani

doc/Manuals/COOLFluiD_FVM_ATD.pdf

@@ -0,0 +1,309 @@
+\documentclass[11pt]{article}
+%*******************************************************************
+%--Packages
+%*******************************************************************
+\usepackage[final]{graphicx,epsfig}
+\usepackage{subfigure}
+\usepackage{subfigmat}
+\newcommand{\noi}{\noindent}
+
+\setlength{\parindent}{0pt}
+\setlength{\parskip}{5pt plus 2pt minus 1 pt}
+\topmargin  -5mm
+\evensidemargin 8mm
+\oddsidemargin  2mm
+\textwidth  158mm
+\textheight 230mm
+%\renewcommand{\baselinestretch}{1.0}
+\frenchspacing
+\sloppy
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+
+\begin{document}
+\pagestyle{empty}
+
+\begin{center}
+  {\fontsize{14}{20}\bf 
+    User-manual for COOLFluiD RDS-Unsteady Navier-Stokes \\
+    (version 2011.04) \\[10pt]}
+\end{center}
+
+\begin{center}
+  {Nad\`{e}ge Villedieu, \underline{[email protected]} \\
+    Von Karman Institute, Aeronautics \& Aerospace Dept.}
+\end{center}
+
+The test case presented here is solving a viscous unsteady flow over a cylinder at $Re=140$ and $M=0.2$. It is set up with adimensionnalized values, which means that the diameter of the cylinder 
+is one and the density and velocity are one. The other parameters are computed such that the Mach and the Reynolds numbers are as expected.
+
+\section{Options that are common to all CFcases}
+
+Necessary modules to run this test case:
+\begin{verbatim}
+Simulator.Modules.Libs = libForwardEuler libCFmeshFileWriter libCFmeshFileReader 
+libTecplotWriter libNavierStokes libFluctSplit libFluctSplitSystem 
+libFluctSplitSpaceTime libFluctSplitSpaceTimeNavierStokes libGambit2CFmesh
+\end{verbatim}
+{\bf WATCH OUT:} The order of the libraries is very important!!! It must be in this order.
+
+
+\subsection{Physical Model}
+The physical model should be set as follows:
+\begin{verbatim}
+Simulator.SubSystem.Default.PhysicalModelType = NavierStokes2D
+
+Simulator.SubSystem.NavierStokes2D.DiffTerm.ViscosityLaw = Constant
+Simulator.SubSystem.NavierStokes2D.DiffTerm.Constant.Value = 0.0000632911
+\end{verbatim}
+The viscosity coefficient is set to a constant value such that the Reynolds number is
+respected.
+
+\subsection{Pseudo-time solver}
+In this case we use an explicit Forward Euler solver for the pseudo-time iterations. The pseudo-time process
+is explained in \cite{AIAA-paper}.
+To choose the type of pseudo-time discretisation the option is:
+\begin{verbatim}
+Simulator.SubSystem.ConvergenceMethod = FwdEuler
+\end{verbatim}
+And the CFL is defined by:
+\begin{verbatim}
+Simulator.SubSystem.FwdEuler.Data.CFL.Value = 0.6
+\end{verbatim}
+{\bf WATCH OUT:} This CFL number is not dictated by the past shield condition (\cite{mr-thesis}).\\
+
+A possible set up of the stop condition of the pseudo-time is:
+\begin{verbatim}
+Simulator.SubSystem.FwdEuler.StopCondition = RelativeNormAndMaxIter
+Simulator.SubSystem.FwdEuler.RelativeNormAndMaxIter.RelativeNorm = -5.0
+Simulator.SubSystem.FwdEuler.RelativeNormAndMaxIter.MaxIter = 500
+\end{verbatim}
+In this case, the (pseudo-time) iterations will stop when the 500 iterations
+has been executed, or when the relative norm of the residual is $-5.0$ (the relative norm is 
+the difference between the residual of the first and of the last iteration). This option is the most convenient,
+but it is also possible to check the absolute norm and the maximal number of iterations (\texttt{AbsoluteNormAndMaxIter}) or
+only the maximal number of iterations (\texttt{MaxNumberSteps}).\\
+
+It is possible choose to print or not the evolution of the residuals:
+\begin{verbatim}
+Simulator.SubSystem.FwdEuler.Data.PrintHistory = false
+\end{verbatim}
+The residuals can be computed using different norms, it is better
+to use L2 norm. Then, it is also possible to define which variable should 
+fufill the stop condition:
+\begin{verbatim}
+Simulator.SubSystem.FwdEuler.Data.NormRes = L2
+Simulator.SubSystem.FwdEuler.Data.L2.MonitoredVarID = 0
+Simulator.SubSystem.FwdEuler.Data.L2.ComputedVarID = 0 1 2 3
+\end{verbatim}
+\subsection{The mesh}
+The mesh is loaded and setup as follow:
+\begin{verbatim}
+Simulator.SubSystem.Default.listTRS = InnerCells wall in out symm_0 symm_1
+Simulator.SubSystem.MeshCreator = CFmeshFileReader
+Simulator.SubSystem.CFmeshFileReader.Data.FileName = kvs2.CFmesh
+\end{verbatim}
+
+\section{RDS}
+\subsection{Solver}
+To choose the residual method as solver:
+\begin{verbatim}
+Simulator.SubSystem.SpaceMethod = FluctuationSplit
+\end{verbatim}
+When one want to restart from a previous solution this option should be on \texttt{true} 
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.Restart = true
+\end{verbatim}
+{\bf WATCH OUT:} The default value of this boolean is \texttt{false}
+
+
+To define which variables are used in each step of the method (to write the solution, to update the 
+residual, to average the solution in each cell in order to compute the Jacobian and to compute the diffusive term)
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.Data.SolutionVar  = Cons
+Simulator.SubSystem.FluctuationSplit.Data.UpdateVar  = Cons
+Simulator.SubSystem.FluctuationSplit.Data.DistribVar = Cons
+Simulator.SubSystem.FluctuationSplit.Data.LinearVar  = Cons
+Simulator.SubSystem.FluctuationSplit.Data.DiffusiveVar = Cons
+\end{verbatim}
+{\bf WATCH OUT:} \textttt{DiffusiveVar} should always be Cons when using \texttt{UnsteadyNavierStokes}. The
+\texttt{LinearVar} are not very important when CRD \cite{mr-thesis} is used to compute the residual. It is better to put it to Roe when
+shocks are considered and N or Nlim schemes are used
+
+To setup the quadrature rule used to compute the residual on the sides of the element:
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.Data.IntegratorQuadrature = GaussLegendre
+Simulator.SubSystem.FluctuationSplit.Data.IntegratorOrder = P3
+\end{verbatim}
+{\bf WATCH OUT:} Those line are extremely important for any system computations because the 
+default quadrature rule is accurate only for linear functions
+
+\subsection{Initialization}
+It is necessary to intialize the field and also the boundaries 
+where a strong condition is used.
+
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.InitComds = InitState \
+ 						 StrongNoSlipWallAdiabaticNS2D 
+
+
+Simulator.SubSystem.FluctuationSplit.InitNames = InField InWall
+
+Simulator.SubSystem.FluctuationSplit.InField.applyTRS = InnerCells
+Simulator.SubSystem.FluctuationSplit.InField.Vars = x y
+Simulator.SubSystem.FluctuationSplit.InField.Def =  1.0 1.0 0.0 350.9087378
+
+Simulator.SubSystem.FluctuationSplit.InWall.applyTRS = wall
+\end{verbatim}
+
+\section{Unsteady RDS for Navier-Stokes}
+
+\subsection{Writing solutions}
+It is possible to change make the name of the output file dependent on the time:
+\begin{verbatim}
+Simulator.SubSystem.CFmesh.AppendTime = true
+Simulator.SubSystem.Tecplot.AppendTime = true
+\end{verbatim}
+
+\subsection{Defenition of the time step}
+There are two possibilities to determine the time step.
+First, one can fix the value of the time step constant 
+over all the computation:
+
+\begin{verbatim}
+Simulator.SubSystem.SubSystemStatus.TimeStep = 0.002
+\end{verbatim}
+{\bf WATCH OUT:} When running in parallel it is mandatory to use this method!!
+
+The second option is to determine the time step at every time iteration,
+then it is computed by $\Delta t= DT \_ Ratio \cdot MaxDT$, where $MaxDT$
+denotes the maximum time step respecting the past shield condition \cite{mr-thesis}
+
+\begin{verbatim}
+Simulator.SubSystem.SubSystemStatus.TimeStep = 0.002
+Simulator.SubSystem.SubSystemStatus.ComputeDT = MaxDT
+Simulator.SubSystem.SubSystemStatus.MaxDT.DT_Ratio = 0.5
+\end{verbatim}
+{\bf WATCH OUT:} The first line is still necessary to fix the first time step. Moreover, this option can be used
+only if the simulation is ran in serial. Indeed, up to now, the time step are not synchronized.
+
+Since there is a restriction on the time step, it is usually interested to know
+what is this limit. So, in general, it is better to first launch one simulation on one processor with two time steps
+and with non fixed time step. Like that it is possible to know what time step should be used. Anyway, in the case of
+a viscous flow, usually, the time step should be smaller that the past shield condition.
+
+\subsection{Stop condition}
+There are two options to stop the computation. The first one is to
+decide to stop the simulation when a threshold time is reached: 
+\begin{verbatim}
+Simulator.SubSystem.StopCondition   = MaxTime
+Simulator.SubSystem.MaxTime.maxTime = 500.0
+\end{verbatim}
+{\bf WATCH OUT:} The time that will be reached is not exactly the one asked. Indeed, the simulation will stop
+as soon as the \texttt{maxTime} is reached or passed: at the last iteration,if the time step is bigger than the difference between the current 
+time and the final one, then the final time will be bigger than \texttt{MaxTime}.
+
+
+The other possibility is to stop the computation when a number of time step
+has been reached
+\begin{verbatim}
+Simulator.SubSystem.StopCondition   = MaxNumberSteps
+Simulator.SubSystem.MaxNumberSteps.nbSteps = 13000
+\end{verbatim}
+
+
+
+\subsection{Solver}
+
+First, we choose what is the strategy used to discretize the unsteady problem.
+The option used here is the space-time where the advective residual is computed using CRD (see \cite{mr-thesis}) 
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.Data.FluctSplitStrategy = STU_CRD	
+\end{verbatim}
+{\bf WATCH OUT:} Only this strategy has been validated for the unsteady Navier-Stokes
+
+The distribution coefficients are defined by the splitter, here we use the space-time version 
+of LDA, two versions are possible \texttt{STKT$\_$SysLDAC} and \texttt{STKS$\_$SysLDAC}, only the first
+one has been validated for the current problem.
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.Data.SysSplitter = STKT_SysLDAC
+\end{verbatim}
+{\bf WATCH OUT:} When starting the simulation from scratch it may be better to start with the N scheme (\texttt{STKT$\_$SysNC}).\\
+
+The command to define the diffusive term is:
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.Data.DiffusiveTerm = UnsteadyNavierStokes
+\end{verbatim}
+
+
+\subsection{Boundary conditions}
+In this case the wall of the cylinder is a non-slipping wall with an adiabatic condition 
+for the energy, the inlet and outlet are strong subsonic conditions which are non-reflecting, and finally, for 
+the top and bottom boundaries we use an hybrid boundary which check if the flow is getting in or out and
+then apply the corresponding conditions.
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.BcComds = StrongNoSlipWallAdiabaticNS2D \
+                                               StrongSubOutletNonRefEuler2DCons \
+                                               StrongSubInletNonRefEuler2DCons \
+						WeakFarFieldEuler2DCons \
+						WeakFarFieldEuler2DCons
+
+Simulator.SubSystem.FluctuationSplit.BcNames = BCWall \
+                                               BCOut \
+                                               BCIn \
+                                               BCSym1 \
+                                               BCSym2
+\end{verbatim}
+
+In the case of the wall and the outlet, only the TRS where the condition should be applied is
+necessary:
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.BCWall.applyTRS = wall
+Simulator.SubSystem.FluctuationSplit.BCOut.applyTRS = out
+\end{verbatim}
+
+For the inlet it is necessary to define the inflow:
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.BCIn.applyTRS = in
+Simulator.SubSystem.FluctuationSplit.BCIn.Vars = x y t
+Simulator.SubSystem.FluctuationSplit.BCIn.InFlow = 1.0 1.0 0.0 350.9087378
+\end{verbatim}
+
+
+For the farfield it is necessary to give the total and static pressure and the total temperature.
+\begin{verbatim}
+Simulator.SubSystem.FluctuationSplit.BCSym1.applyTRS = symm_0
+Simulator.SubSystem.FluctuationSplit.BCSym1.Ttot = 0.488793964
+Simulator.SubSystem.FluctuationSplit.BCSym1.Ptot = 140.6641324
+Simulator.SubSystem.FluctuationSplit.BCSym1.P = 140.1634951
+
+Simulator.SubSystem.FluctuationSplit.BCSym2.applyTRS = symm_1
+Simulator.SubSystem.FluctuationSplit.BCSym2.Ttot = 0.488793964
+Simulator.SubSystem.FluctuationSplit.BCSym2.Ptot = 140.6641324
+Simulator.SubSystem.FluctuationSplit.BCSym2.P = 140.1634951
+\end{verbatim}
+
+
+\subsection{Post-processing}
+It is possible to write in a file the evolution of the solution in one point:
+\begin{verbatim}
+Simulator.SubSystem.DataPostProcessing = DataProcessing
+Simulator.SubSystem.DataPostProcessingNames = DataProcessing2
+Simulator.SubSystem.DataProcessing2.ProcessRate = 1
+Simulator.SubSystem.DataProcessing2.Comds = SamplingPoint
+Simulator.SubSystem.DataProcessing2.Names = SPoint
+Simulator.SubSystem.DataProcessing2.SPoint.OutputFile = samplingbis.plt
+Simulator.SubSystem.DataProcessing2.SPoint.location = 1.0 0.0
+Simulator.SubSystem.DataProcessing2.SPoint.SaveRate = 1
+\end{verbatim}
+
+
+
+\begin{thebibliography}{02}
+\bibitem{AIAA-paper} L. Koloszar ,N. Villedieu, T. Quintino, J. Anthoine and P. Rambaud. Application of Residual Distribution Method for Acoustic Wave Propagation.
+{\emph{AIAA journal}}, AIAA 2009-3116 (2009)
+\bibitem{mr-thesis} M. Ricchiuto. {\emph{Construction and Analysis of Compact Residual Discretizations for Conservation Laws on Unstructured Meshes}}. Universit{\'e} Libre de Bruxelles (2005),
+ISBN: 2-930389-16-8
+\end{thebibliography}
+
+\end{document}