Feature CFL adaptation initial iteration (#2362)

* Feature to define initial iteration to apply CFL adaption.

A spark initialization was not possible with CFL adaptation, as the CFL number would increase too far before the spark initiates the combustion. This led to immediate flashback or blow-off. An additional parameter in the CFL adaptation defines the iteration, at which the CFL adaption should start. This allows the combustion to stabilize before the CFL number increases.

Co-authored-by: Nijso <[email protected]>
Co-authored-by: TobiKattmann <[email protected]>
Co-authored-by: Nijso <[email protected]>
Co-authored-by: Pedro Gomes <[email protected]>

Common/include/CConfig.hpp

@@ -1109,7 +1109,7 @@ class CConfig {
   bool Radiation;                      /*!< \brief Determines if a radiation model is incorporated. */
   su2double CFL_Rad;                   /*!< \brief CFL Number for the radiation solver. */
 
-  array<su2double,5> default_cfl_adapt;  /*!< \brief Default CFL adapt param array for the COption class. */
+  array<su2double,6> default_cfl_adapt;  /*!< \brief Default CFL adapt param array for the COption class. */
   su2double vel_init[3], /*!< \brief initial velocity array for the COption class. */
   vel_inf[3],            /*!< \brief freestream velocity array for the COption class. */
   eng_cyl[7],            /*!< \brief engine box array for the COption class. */

Common/src/CConfig.cpp

@@ -1777,11 +1777,16 @@ void CConfig::SetConfig_Options() {
   /* DESCRIPTION: Activate The adaptive CFL number. */
   addBoolOption("CFL_ADAPT", CFL_Adapt, false);
   /* !\brief CFL_ADAPT_PARAM
-   * DESCRIPTION: Parameters of the adaptive CFL number (factor down, factor up, CFL limit (min and max), acceptable linear residual )
+   * DESCRIPTION: Parameters of the adaptive CFL number (factor down, factor up, CFL limit (min and max)[, acceptable linear residual][, starting iteration]).
+   * Parameters in square brackets are optional, parameter "starting iteration" only valid with parameter "acceptable linear residual".
    * Factor down generally <1.0, factor up generally > 1.0 to cause the CFL to increase when the under-relaxation parameter is 1.0
    * and to decrease when the under-relaxation parameter is less than 0.1. Factor is multiplicative. \ingroup Config*/
-  default_cfl_adapt[0] = 1.0; default_cfl_adapt[1] = 1.0; default_cfl_adapt[2] = 10.0; default_cfl_adapt[3] = 100.0;
+  default_cfl_adapt[0] = 0.1;
+  default_cfl_adapt[1] = 1.2;
+  default_cfl_adapt[2] = 10.0;
+  default_cfl_adapt[3] = 100.0;
   default_cfl_adapt[4] = 0.001;
+  default_cfl_adapt[5] = 0.0;
   addDoubleListOption("CFL_ADAPT_PARAM", nCFL_AdaptParam, CFL_AdaptParam);
   /* DESCRIPTION: Reduction factor of the CFL coefficient in the adjoint problem */
   addDoubleOption("CFL_REDUCTION_ADJFLOW", CFLRedCoeff_AdjFlow, 0.8);
@@ -7230,7 +7235,8 @@ void CConfig::SetOutput(SU2_COMPONENT val_software, unsigned short val_izone) {
       if (!CFL_Adapt) cout << "No CFL adaptation." << endl;
       else cout << "CFL adaptation. Factor down: "<< CFL_AdaptParam[0] <<", factor up: "<< CFL_AdaptParam[1]
         <<",\n                lower limit: "<< CFL_AdaptParam[2] <<", upper limit: " << CFL_AdaptParam[3]
-        <<",\n                acceptable linear residual: "<< CFL_AdaptParam[4] << "." << endl;
+        <<",\n                acceptable linear residual: "<< CFL_AdaptParam[4]
+        <<"'\n                starting iteration: "<< CFL_AdaptParam[5] << "." << endl;
 
       if (nMGLevels !=0) {
         PrintingToolbox::CTablePrinter MGTable(&std::cout);

SU2_CFD/src/solvers/CSolver.cpp

@@ -1717,6 +1717,7 @@ void CSolver::AdaptCFLNumber(CGeometry **geometry,
   const su2double CFLMin            = config->GetCFL_AdaptParam(2);
   const su2double CFLMax            = config->GetCFL_AdaptParam(3);
   const su2double acceptableLinTol  = config->GetCFL_AdaptParam(4);
+  const su2double startingIter      = config->GetCFL_AdaptParam(5);
   const bool fullComms              = (config->GetComm_Level() == COMM_FULL);
 
   /* Number of iterations considered to check for stagnation. */
@@ -1764,8 +1765,12 @@ void CSolver::AdaptCFLNumber(CGeometry **geometry,
 
     /* Check if we should decrease or if we can increase, the 20% is to avoid flip-flopping. */
     resetCFL = linRes > 0.99;
-    reduceCFL = linRes > 1.2*linTol;
-    canIncrease = linRes < linTol;
+    unsigned long iter = config->GetMultizone_Problem() ? config->GetOuterIter() : config->GetInnerIter();
+
+    /* only change CFL number when larger than starting iteration */
+    reduceCFL = (linRes > 1.2*linTol) && (iter >= startingIter);
+
+    canIncrease = (linRes < linTol) && (iter >= startingIter);
 
     if ((iMesh == MESH_0) && (Res_Count > 0)) {
       Old_Func = New_Func;

config_template.cfg

@@ -1389,12 +1389,12 @@ CFL_NUMBER= 15.0
 CFL_ADAPT= NO
 %
 % Parameters of the adaptive CFL number (factor-down, factor-up, CFL min value,
-%                                        CFL max value, acceptable linear solver convergence)
-% Local CFL increases by factor-up until max if the solution rate of change is not limited,
-% and acceptable linear convergence is achieved. It is reduced if rate is limited, or if there
+%                                        CFL max value, acceptable linear solver convergence, starting iteration)
+% After the starting iteration has passed, local CFL increases by factor-up until max if the solution rate of
+% change is not limited, and acceptable linear convergence is achieved. It is reduced if rate is limited, or if there
 % is not enough linear convergence, or if the nonlinear residuals are stagnant and oscillatory.
 % It is reset back to min when linear solvers diverge, or if nonlinear residuals increase too much.
-CFL_ADAPT_PARAM= ( 0.1, 2.0, 10.0, 1e10, 0.001 )
+CFL_ADAPT_PARAM= ( 0.1, 2.0, 10.0, 1e10, 0.001, 0)
 %
 % Maximum Delta Time in local time stepping simulations
 MAX_DELTA_TIME= 1E6