DEBUG_GeneralParameter

General list of debug parameters at the developers disposal

for development only
DEBUG_GeneralParameter = (1.0, 2.0, 3.0, ...)
Default: DEBUG_GeneralParameter = 0 Currently involved:
  • DEBUG_GeneralParameter(1)...DEBUG_GeneralParameter(4) :: testing for DIFFOP_Version=10

  • DEBUG_GeneralParameter(5) :: testing COMP_GradtEtaGrad_Version=%GradtEtaGrad_Identity%

  • DEBUG_GeneralParameter(6) = 3 (default=2) :: chose the way how INTERPHASE_f-operator is established 1::pposite point by its localapproximation 2::opposite point by delta function 3:: opposite point by Shepard-approximation

  • DEBUG_GeneralParameter(7) = -1 :: switch off special velocity correction at free surfaces

  • DEBUG_GeneralParameter(8) = -1 :: switch off special re-interpolation of newly created free surface points

  • DEBUG_GeneralParameter(9) = 1.5 (default=1000) :: DIFFOP_Version=9: for point "i", step back from order=3 to order=2 if \( \max \limits_{j=1...N(i)}\left( c_j^x (x_j-x_i) + c_j^y (y_j-y_i) + c_j^z (z_j-z_i) \right) > 1.5\)

  • DEBUG_GeneralParameter(9) = 0.7 (default=1000) :: DIFFOP_Version=91: for point "i", go down one approximation order if difference between this order and one lower order is bigger than this value

  • DEBUG_GeneralParameter(10) = 0.5 (default=1000) :: DIFFOP_Version=9: for point "i", step back from order=2 to order=1 if \( \max \limits_{j=1...N(i)}\left( c_j^x (x_j-x_i) + c_j^y (y_j-y_i) + c_j^z (z_j-z_i) \right) > 0.5\)

  • DEBUG_GeneralParameter(10) = 0.7 (default=1000) :: DIFFOP_Version=91: for point "i", go down one approximation order if difference between this order and one lower order is bigger than this value

  • DEBUG_GeneralParameter(11) = 0.1 (default=0.0) :: Smagorinsky-Lilly-ansatz for viscosity (SLA) in degenerated phases/films: \( \eta_{film} = \max\left( \eta^{*} , \rho (C_{SLA} D)^2 \frac{\|\mathbf{v}\|}{D} \right)\), with \( D =\) film thickness, if \( C_{SLA}<0\) then \( \eta^{*} =\eta_{turblent}\), else \( \eta^{*} =\eta_{laminar}\)

  • DEBUG_GeneralParameter(12) = 1.5 (default=1000) :: DIFFOP_Version=9: safety threshold for interior points (TODO Tobias)

  • DEBUG_GeneralParameter(13) = 0.05 (default=0.1) :: DIFFOP_Version=9: numerical differentiation step size D (relative to h): c_x = (c_0(x+D)-c_0(x-D))/(2*D), step size D=0.1 (default), can be adapted by this parameter

  • DEBUG_GeneralParameter(14) = 8 (default=3) :: DIFFOP_Version=9: interior points: drop from order 3 to order 2, if norm(cx,cy,cz) > DEBUG_GeneralParameter(14), default: 3.0

  • DEBUG_GeneralParameter(15) = 25 (default=16) :: DIFFOP_Version=9: boundary points: drop from order 3 to order 2, if norm(cx,cy,cz) > DEBUG_GeneralParameter(15), default: 16.0

  • DEBUG_GeneralParameter(16) = 0.1 (default=1.0) :: free surface boundary conditions on velocity: weight for the div(v)=rhs condition, default: 1.0

  • DEBUG_GeneralParameter(17) = 0.0 (default=1.0) :: free surface boundary conditions on velocity: weight for the div(v)=alpha*rhs condition, where DEBUG_GeneralParameter(17) describes the value of alpha, default: 1.0

  • DEBUG_GeneralParameter(18) = 0.0 (default=3.0) :: Shepard-smoothing-coefficient of \( \frac{ d{\bf v} }{dt}\) if second digit of FLIQUID_ConsistentPressure_Version is 2

  • DEBUG_GeneralParameter(19) = 1.0 (default=0.0) :: switch on the over-weighting of boundary points for establishing differential operators, in order to hope for better stability in case of locally bad point cloud quality. The weight is usually 1, and given in %ind_vol%. Important: with this parameter, one has to explicitly SWITCH ON, what was switched on by default in all version until beta2020.10.2

  • DEBUG_GeneralParameter(20) = 1.0 (default=0.3) :: switch on the boundary normal adaption in case of bad quality of the local point cloud (if Y%ind_EdgeValue% > DEBUG_GeneralParameter(20)), \begin{align} \mathbf{n}_{i,\text{adapted}} = \frac{ \mathbf{x}_{i,\text{COG}} - \mathbf{x}_i }{ \| \mathbf{x}_{i,\text{COG}} - \mathbf{x}_i \|}\end{align} This might help to gain stability of the computations (however, there is no proof, that it helps in any case). If using a negative number, then the adapted normal IS NOT copied to Y%ind_n(1)% ... Y%ind_n(3)%, BUT ONLY used for establishing the differential Neumann operator \( \frac{\partial}{\partial n}\). and then one would run into trouble for example in the Ticket 88813 (add blue tank)

  • DEBUG_GeneralParameter(21) = 1.0 (default=0.0) :: switch on the additional term \( \nabla^T \left( {\mathbf g} - \frac{1}{\rho} \nabla p_{hyd} \right)\) on the right hand side of DynamicPressureAlgorithm -> ClassicalDPA . Theoretically, this value is strictly zero (after solving HydrostaticPressureAlgorithm) , however, numerically, it is not exactly zero. Thus, with this term in action, one would correct the numerical errors from the hydrostatic pressure solution.

  • DEBUG_GeneralParameter(22) = 0.0 (default=1.0) :: switch on rejection of velocity correction by Chorin, if ONE point has a critical correction value \( -\frac{\Delta t_{virt}}{\left( 1+\Delta t_{\beta }\cdot \beta \right)}\frac{1}{\rho }\nabla c\), see CorrectionPressureAlgorithm . This stability feature was active up to beta2020.11.0 , but seems to be counter-productive and is harmful to some of the watercrossing applications, tank slohsing and filling.

  • DEBUG_GeneralParameter(23) = 0.0 (default=1.0) :: switch off/on improvements of pointcloud organization: do not allow boundary points to fill interior points, if Y%ind_EdgeValue% > 0.3 or Y%ind_act% < 2 . That avoid failure-filling in regions where geometry slightly intersects

  • DEBUG_GeneralParameter(24) = 0.0 (default=3.0) :: switch off/on improvements of pointcloud organization: if dist_LayerThickness is given, keep the distance of moving %BND_free%-points already during point displacement (previously, the boundary distance was corrected in the distance-to-boundray-function. This provides a better/safer handling of point-through-boundary-penetrations in the vicinity of slightly intersecting geometries. == 1 :: correct location only == 2 :: correct location AND velocity (restrict velocity by a constraint on the kinetic energy) == 3 :: correct location AND velocity (velocity component normal to the wall equal normal wall velocity)
  • DEBUG_GeneralParameter(25) = 0.0 (default=0.3) :: RENEW HELP ARRAY for SMOOTH_LENGTH : speed up by partially omitting/excluding very small triangles from the loop. Consider triangle as candidate for omitting if DL < DEBUG_GeneralParameter(25)*max( L_min, h_min ) h_min and L_min given by the definition of SMOOTH_LENGTH , such as %H_spherical% or %H_linear% DL = sqrt( triangleArea ) re-establish old behavior: set this value to ZERO, i.e. DEBUG_GeneralParameter(25) = 0.0
  • DEBUG_GeneralParameter(26) = 0.0 (default=0.1) :: RENEW HELP ARRAY for SMOOTH_LENGTH : re-consider triangle, if previously triangles were omitted, for which we have sum( triangleArea_i ) > DEBUG_GeneralParameter(26)*( max( L_min, h_min )^2 ) re-establish old behavior: set this value to ZERO, i.e. DEBUG_GeneralParameter(26) = 0.0
  • DEBUG_GeneralParameter(27) = 1.0 (default=0.0) :: by default, for solving the vp- system (using BiCG), the initial guess for the correction pressure is 0. Use the solution of %ind_c% of the previous time step as initial guess.

  • DEBUG_GeneralParameter(28) = 10.0 (default=1.0) :: scale up/down %ind_PenV% by this factor (for convergence studies)

  • DEBUG_GeneralParameter(29) = 10.0 (default=1.0) :: scale up/down %ind_penalty% by this factor (for convergence studies)

  • DEBUG_GeneralParameter(30) = 10 (default=0.2) :: switch off all quality parameters for the laplace-operator. If DEBUG_GeneralParameter(30)=1, then only switch off the first one. If < 1, then notBremse is activated if Y%indEdgeValue% < DEBUG_GeneralParameter(30)

  • DEBUG_GeneralParameter(31) = 1.0 (default=0.0) :: computation of PHI: number of smoothing loops

  • DEBUG_GeneralParameter(32) = 1.0 (default=3.0) :: computation of PHI: coeff for smoothing kernel

  • DEBUG_GeneralParameter(33) = 2.0 (default=1.0) :: computation of PHI: correction item is Dind_PHI** DEBUG_GeneralParameter(33)

  • DEBUG_GeneralParameter(34) = 1.0 (default=0.0) :: computation of PHI: scale the computed required INT(PHI)

  • DEBUG_GeneralParameter(35) = 1.0 (default=0.0) :: switch on sophisticated point transport in case of EULER

  • DEBUG_GeneralParameter(36) = 0.5 (default=0.0) :: additional condition on the Neumann operator of 3rd order: if > 0 :: define the result for the operator acting on (xn^2), the analytical value would be 0, but leads to unstable operators if < 0 :: define the number of ansatz functions. Currently, only -9 is allowed.
  • DEBUG_GeneralParameter(37) = 3 (default=10000) :: IN CASE OF v-- , adapt the local value of \( \tilde{ \Delta t_\beta } \left( \mathbf{x}_i \right)\) ( see CorrectionPressureAlgorithm ) such that \begin{align} \tilde{ \Delta t_\beta }\left( \mathbf{x}_i \right) \cdot \beta_i \le \text{DGP}(37) \cdot \max\left( 1 , \frac{ \Delta t \eta_i }{\rho_i H_i^2} \right)\end{align} which comes from the assumtion, that the term \( \frac{\tilde{ \Delta t_\beta } \beta H}{U_0 \Delta t }\) must not be major to the Strouhal number \( \frac{H}{U_0 \Delta t}\) or the Reynolds number \( \frac{\eta}{\rho H U_0}\) The final value of \( \Delta t_\beta\) for the CorrectionPressureAlgorithm is given by additional application of COEFF_dt_Darcy on top of it: \( \Delta t_\beta = \text{COEFFdtDarcy} \cdot \tilde{ \Delta t_\beta }\)
  • DEBUG_GeneralParameter(38) = 2 (default=3) :: set minimum trial-order for Laplacian differential operators if minimum trial order fails to meet the quality criteria, then the socalled "Notbremse" (emergency break) is applied as differential operator \( c_{ij}^\Delta=\frac{8}{h_i^2}\frac{1}{N_i-1}\) and \( c_{ii}^\Delta = -\frac{8}{h_i^2}\) which is ultimately stable, has typical Laplacian smoothing/damping qualities, but DOES NOT provide any approximation order consitency, except that it projects constant functions identically to zero. \( N_i\) is the numebr of active neighbors to the point $i$, including the central point itself.
  • DEBUG_GeneralParameter(40) = 6 (default=10000) :: for points detaching from wall (edge-detachment), this is the iMethod-index to check whether the detached point is %BND_free% or %BND_none% (if this parameter is not given, then use second digit of ORGANIZE_CheckFreeSurface_Version) if 10000 then the detached point becomes %BND_none%, and the next regular free-surface-check naturally decides. This is only done in LIQUID chambers. In order to invoke the old behavior, actively set DEBUG_GeneralParameter(40) = -1
  • DEBUG_GeneralParameter(41) = 2 (default=0) :: number of smoothing loops for the values of PSI and THETA ( right hand side values for the DynamicPressureAlgorithm )
  • DEBUG_GeneralParameter(42) = 0 (default=-0.0001) :: switch on/off the additional source terms of G at wall points in the k-epsilon model due to %ind_TurbulentWallLayer% \( G_\text{adapted} = (1.0-\gamma) \cdot G_\text{orig} + \gamma \frac{ \left( \mathbf{v}-\mathbf{v}_p \right)^2 }{ \left( \alpha H \right)^2 }\) where \( \gamma = \text{DBG(42)}\) where \( G\) is the square norm of the velocity gradient as used for the production terms in the k-epsilon model, see DOCUMATH_NumericalIntegrationOfTurbulence.pdf , \( P_k\), and \( \alpha H\) is the thickness of the wall layer, i.e. Y%ind_TurbulentWallLayer% * Y%ind_h% . NEW !!! :: setting this parameter <= 0 , then MESHFREE computes the weight on its own in the sense \( \gamma = \frac{\eta \cdot \left| \mathbf{v}-\mathbf{v}_p \right| }{ \tau_W \cdot \beta H }\) , with \( \gamma > | \text{DBG(42)} |\) , see also the document. Note: \( \alpha\) = %ind_turbulentWallLayer% and \( \beta\) = momentume thickness as usual (see in the document mentioned/linked above)
  • DEBUG_GeneralParameter(49) = 0.2 (default=0.1) :: set relative stecil-distance for which to reduce the weight of the neighbor CAREFUL: for wall points, dist_LayerThickness will dominate this value: DBG(49) := min( DBG(49) , dist_LayerThickness ) CAREFUL: for interior points, dist_rip will dominate this value DBG(49) := min( DBG(49) , dist_rip )
  • DEBUG_GeneralParameter(50) = 0.01 (default=0.1) :: set factor by which to reduce the weight of the neighbor if closer that the threshold given in DBG(49)
  • DEBUG_GeneralParameter(51) = 0.0 (default=1.0) :: switches on the term \( \frac{\partial }{\partial m}\) in the %BND_slip% boundary condition, see section 2.2 in DOCUMATH_LIQUID_BoundaryConditions.pdf
  • DEBUG_GeneralParameter(52) = 10.0 (default=0) :: determines how to actually compute the transport operator in case of BC_v($BC$) = ( %BND_slip%, ... ) in case of EULER or EULERIMPL option 1) if DBG(52) == 0 use the native transport operator of the running scheme (such as EULERIMPL, EULER) option 2a) if DBG(52) == 10 :: new formulation with explictly computing the transport operator 2nd order UPWIND -->> The new idea option 2b) if DBG(52) == 11 :: same as 10, however reduce the ansatz functions to those being in the plane of the boundary option 2c) if DBG(52) == 12 :: same as 11, however increase weight of boundary points by factor 10 option 2d) if DBG(52) == 13 :: same as 11, however increase weight of boundary points by factor 100 option 3) if DBG(52) == 1 (or 2, 3, 4) :: old formulation transport = gradient + DBG(52)*delta_x*diffusion(very rough) option 4) if DBG(52) == -0.5(or -1.0) :: old formulation transport = gradient + abs(DBG(52))*delta_x*diffusion(reduced number ansatz functions) -0.5: second space-time-order, -1.0: first space-time-order
  • DEBUG_GeneralParameter(53) = 0.01 (default=10) :: transition of the dt_virt for y_star=0 up to y_star=11.225 ((y_star/11.225)**DBG(53) ), see %TAUW_PressureDependent%
  • DEBUG_GeneralParameter(54) = 1.0e-7 (default=1.0e-40) :: define minimal turbulent k, that is applied in KEPSILON_Repair-functinalities if a MINUS-sign is put in front, then use the KEPSILON_Repair_2-functionality (that is use the projection of the values of the healthy points) HINT: repair of k-epsilon takes place, if the value of k is Nan, Inf, or below the given value. In ths case, epsilon is set such that the change rate of k becomes zero.
  • DEBUG_GeneralParameter(55) = 1 (default=0) :: modifications for the K-EPSILON-solver in case of EULERIMPL 1 -> call FLIQUID_KEPSILON(original Lagrange-implementation) and add EULERIMPL-space-discretization (such as MUSCLE-reconstruction) for the transport part 0 -> original EULERIMPL-implementation (such as MUSCLE together with SDIRK2 etc) -1 -> original implementation, BUT downgrade the SDIRK2-agorithm (if given) to upwind
  • DEBUG_GeneralParameter(56) = 0.1 (default=0.3) :: modify pushing of K and EPSILON (description on pushing will follow in the k-epsilon document) after pushing, the turbulent viscosity has to be DEBUG_GeneralParameter(56)*Smagorinsky-viscosity Hint: pushing takes place if gradient of velocity suggests a high change rate of k, however the current value of k is too small Hint: the 0.3 seems to be the best value after intensively studying turbulent boundary layers
  • DEBUG_GeneralParameter(57) = 2 (default=0) :: modifications on the pushing of K and EPSILON 0 :: leave epsilon untouched, i.e. the turbulent viscosity is DEBUG_GeneralParameter(56)*nue_Smagorinskij 1 :: chose epsilon such that the change rate of k becomes zero 2 :: chose epsilon such that the change rate of nue_turb becomes zero 3 :: chose epsilon such that the change rate of (k/eps) becomes zero
  • DEBUG_GeneralParameter(58) = 1 (default=0) :: modify the solution of the momentum equation (solve_V) in case of EULERIMPL 0 -> original space discretization (MUSCLE etc.) with SDIRK2 (or upwind) 1 -> use the LAGRANGE implementation and add the EULERIMPL-space-discretization as transport term (such as MUSCLE)
  • DEBUG_GeneralParameter(59) = 0.2 (default=0.05) :: minimum shift for the upwind-transport operator in case of EULER L_shift = max( DBG(59)*h , v_Euler*dt )
  • DEBUG_GeneralParameter(60) = 90 (default=0) :: how to establish the upwind -transport operator in case of EULER . 0: du/dm = ( u(x) - u(x-Lshift*m) )/Lshift (m is the transport direction, . u is the least equares approximation at the upwind position) . u is solved exactly for the basic monomials 1, x, y, z, x^2 . >= 1: du/dm = m * grad(u)(x-0.5*Lshift*m) . exact transport solution requested for monomials 1, x, x^2 for given time step size . additionally, request exact transport for "delta"-function exp( -A*x^2 ), NOTE A=DBG(60)
  • DEBUG_GeneralParameter(61) = 1.4 (default=1.0) :: kappa value ( \( \kappa\) ) for the bubble pressure computation: \( p_{bubble} \cdot V_{bubble}^{\kappa} = \text{const}\)
  • DEBUG_GeneralParameter(62) = 0.01 (default=0.0) :: resolve the difference between true bubble pressure and Y%ind_pBubble% :: 0.01 try to resolve 1 percent of the difference per time step.
  • DEBUG_GeneralParameter(63) = 0.1 (default=0.05) :: locked points: factor of reduction of time step size subcycling compared to the contact time . all other time integration remains on the current MESHFREE-timestep
  • DEBUG_GeneralParameter(74) = 0.01 :: EXPERIMENTAL in case of EULERIMPL, define the time step size for the velocity integration with the SDIRK2-algorithm. . all other time integration remains on the current MESHFREE-timestep . Remark : sometimes, the SDIRK2-iteration leads to exponential growth of the velovity, when dt(EULERIMPL)>>dt(LAGRANGE) AND the streamlines have . a curvature radius in the order of the local SMOOTH_LENGTH . . In this case, it is legitim to virtually reduce the time step size for the velocity integration (preferably to the size of LAGRANGE), if a quasi-stationary velocity solution can be assumed.
  • DEBUG_GeneralParameter(75) = 1.0 :: Performance Measurement CC2 : 0.0 OFF (default), 1.0 (On)
  • DEBUG_GeneralParameter(77) = 0.0 :: MPI Optimization for Allreduce : 0.0 no optimization, 1.0 optimized (default), 2.0 turned off
  • DEBUG_GeneralParameter(78) = 0.0 (default=2.0) :: TEMPORARY PARAMETER ! Will be moved to RepresentativeMass_iData , soon. In the DefinitionRepresentativeDensity , we add the possibility to smooth the representative density in the sense \( \overset{\scriptscriptstyle\frown}{\rho }_{i}^{{smooth}} = \frac{ \sum_j W_{ij} \overset{\scriptscriptstyle\frown}{\rho }_{j} }{ \sum_j W_{ij} }\) This parameter shall be an integer value and defines, how often this smoothing is repeated.
  • DEBUG_GeneralParameter(79) = 0.0 (default=2.0) :: TEMPORARY PARAMETER ! . Will be moved to RepresentativeMass_iData , soon. In the DefinitionRepresentativeDensity , additional smoothing is invoked with DEBUG_GeneralParameter(78) This parameter defines the \( \alpha\) in the definition of the weight function \( W_{ij} = exp( -\alpha \frac{ r_{ij}^2}{ h_{ij}^2 } )\)
  • DEBUG_GeneralParameter(82) = 1.0 (default=0) :: switch on using Y%ind_tauWv(1)% ... Y%ind_tauW_v(3)% for the computation of Y%ind_Sn(1)% ... Y%ind_Sn(3)% In the default case, it is using Y%ind_tauW% in the direction of the relative velocity.
  • DEBUG_GeneralParameter(83) = -0.5 (default=-0.8) :: reject neighbor for assign_dA if min( -x_ij*n_i, x_ij*n_j ) < DEBUG_GeneralParameter(83)
  • DEBUG_GeneralParameter(84) = 0.3 (default=0.2) :: bubble rejection if max(sum(n*dA)) > DEBUG_GeneralParameter(84)*sum(dA) ; remember that int(n*dA)=0 for closed surfaces in theory.
  • DEBUG_GeneralParameter(85) = 1.0 (default=0.0) :: Improve the internal pressures generated in the hydrostatic pressure algorithm esp. when the plastic shear modulus falls drastically.
  • DEBUG_GeneralParameter(86) = 1 (default=0) :: 1 -> then do NOT do liquid layer analysis 2 -> then do NOT do liquid layer analysis, also NO tear off of the point based on gravity force and surface tension
  • DEBUG_GeneralParameter(87) = 0 (default=1) :: use old version of giving dummy values to non initialized acronyms ($...$-variables). 0 -> old version of giving positive values, however then occasionally we run into troubles in case of undefined-but-used equations, see ticket #102082 1 -> give negative values :: the code might occasionally break down for non-defined material acronyms (the user will then have to make consistent the input file)
  • DEBUG_GeneralParameter(88) = 0 (default=1) :: 1 -> MPI optimized version in saving ERF 0 -> old version of MPI communication in HDF5ERF
  • DEBUG_GeneralParameter(89) = 0 (default=1) :: 1 -> ORGANIZE_DetectEdgesInBE is recomputed every time BND search tree is affected. 0 -> ORGANIZE_DetectEdgesInBE is computed only in the first timestep only in the FSI context.
  • DEBUG_GeneralParameter(90) = 0.001 (default=0) :: 0.001 -> Any positive value will be the absolute length of all BEnormals 0 -> Any value less than or equal to 0 will scale the BEnormals based on the BEarea. Currently this value is 0.3*BEarea()
  • DEBUG_GeneralParameter(92) = 0.0 (default=1.0) :: SAVE_BE_filter: 0.0 => V1, 1.0 => V2, 2.0 => V2 + debugging output
  • DEBUG_GeneralParameter(94) = 0 (default=1) :: switch off the performance speedup at MESHFREE-startup rewgarding loops in BE-environment. The values mean
This item is referenced in:
k-epsilon_development control recent developments of the k-epsilon model
DEBUG_GeneralParameter General list of debug parameters at the developers disposal
DIFFOP_Version version of least squares operators
ord_convection define approximation order of the transport operators
COEFF_dt_Darcy define the virtual time step size for applications with Darcy (Brinkman) term (UCV)
Beta Latest release notes for the MESHFREE beta executables
All Complete release notes for the MESHFREE beta executables
DefinitionRepresentativeDensity define the representative density