Simulate with complex geometries and complex physics

Beta

Latest release notes for the MESHFREE beta executables

Short URL to this page: https://meshfree.pages.fraunhofer.de/docu/beta

WARNING: beta versions -- reduced test suite only

If you would like to be notified about new beta versions of MESHFREE, please send an email to meshfree-beta-subscribe@itwm.fraunhofer.de (or ask our Support team for help).

beta2026.03.0

New: Tickets: Bugfixes: Known Limitations:
  • In some cases, there is an MPI deadlock. Workaround options:
    • Turn off shared memory with COMP_SharedMemoryForBE = false in common_variables.dat
    • Ensure to run the simulation on the same number of MPI processes per node

beta2026.01.0

New:
  • BUBBLE_PressureDrop: new functionality to reduce the pressure inside of bubbles if needed.
  • BUBBLE_DoTheManagement: the 6-th digit can now be used to avoid rejection of bubbles that touch %IDENT_outflow% boundaries. In this case, BUBBLE_PressureDrop can be used to let gas (naturally) ventilate out of the bubble with an appropriate pressure drop.
  • CODI_min_max / ENFORCE_min_max : new functionality of restricting the number of iteration loops. Iteration loops will be necessary if working with big numbers of MPI processes.
  • STANDBY pointclouds : now there are options for using EVENTS during read-in of pointclouds using begin_pointcloud{} environment
  • new functionality in locop( %EQN_locop_gapsize%, ... ) : now, the last optional values 12,13,14 can be used to re-define the launch point of the ray.
  • ReconstructionOfSmoothLength : allows to provide options for the smoothing length reconstruction on the STANDBY-PC
  • ReadingENSIGHTGOLD : allows to provide options for reading ENSIGHT GOLD data. Due to the new options, the reading of ENSIGHT GOLD considerably speeds up und uses much less temporary memory.
Tickets:
  • #114567: Improved error handling of SAMG.
  • #114288: Allow locked points for parceling. Caution: parceling between locked points is now the default. If this is not desired it has to be deactivated with DROPLETPHASE_AddParcelingCond, see Parcels.
Bugfixes:
  • Fixed problem in ORGANIZE_OppositePoints_Version = 3.
  • Bugfix for boundary conditions of %ind_c% in conjunction with %BND_free_implicit% and the BubbleAlgorithm.
  • STRESSTENSOR_Variante_Factor : initialize with 50 (before, initialization was at a value of zero, which was physical and numerical nonsense)
  • if working with STANDBY pointclouds, the Bisection was leading to partially wrong values in %ind_proc%
  • damping_p_corr was not effective in case of PDYN:NONE . This bug was in the code since beta2025.06.0 and is now fixed.

beta2025.11.0

New: Tickets:
  • #113750: there was a time lapse of the processor computation (%ind_proc%) between regular pointcloud (LIQUID, DROPLETPHASE, GASDYN, ... ) and STANDBY pointcloud. After a new MPI_Bisection, the refreshment of %ind_proc% did not happen synchronously. That led to wrong interpolation results when mapping between regular and STANDBY pointclouds.
Bugfixes:
  • MPI_CommunicationMethod default reverted to 2, due to repeatedly observed error exits with previous default of 4.
  • For RIGIDBODY_TimeIntegrationVersion == 1, gravity is now computed even if a rigid body is not in contact with the pointcloud. In this context, the default of RIGIDBODY_TimeIntegrationVersion is now set to 2 (version 1 is rather outdated)
  • After any MPI_bisection, there was missing a subsequent MPI-processor recomputation ( %ind_proc% ) in the same time step. That led to the problem of ticket #113750 (see above)

beta2025.09.0

New:
  • EVENT: can now be applied also to STANDBY pointclouds using the keyword %EVENT_only_PC_STANDBY%.
  • The default number of days an encrypted file remains valid has been changed from 3000 to 100. Details how to set the number of days, an encrypted file remains valid, can be found under Encryption.
  • SAMG: Upgraded to version 2024.6; errors in small components do not cause the overall method to fail.
Tickets:

beta2025.08.1

Tickets:

beta2025.08.0

New:
  • VOLUME_correction_TimeScale: a new means of control / stabilization for the volume correction such as VOLUME_correction_local
  • DtDom(): introduced a new second argument, %OnlyCheckLines%, to the DtDom() functionality. This enables selective checking of line segments only.
  • MassSink (experimental): new keyword is introduced to manipulate PC-indices for points located close to geometrical gaps.
  • When USER_h_PC_refinement is active, version 3 of NEIGHBOR_FilterMethod is skipped.
  • Removed common_variables parameter MESHFREE_LICENSE_FILE, which became obsolete with the switch to FlexLM in beta2024.09.1. See InstallationGuide for details on how to use licenses.
  • Added an option to provide absolute shift of wallpoints in the normal direction in the NEIGHBOR_FilterMethod, see Version=3+BE.
  • New CommandLine option --write-ucv to write Ucvlines during startup (debugging purposes), see CONTROL_writeUcvLines
  • New CommandLine option --educational to use limited educational license. See LicenseVariants.
  • FORCE:DUMMYNODE and FORCE:DUMMYPOINT two new options to force a dummy point in the ENSIGHT6 BINARY results in order to avoid empty pointclouds in the case-files. See SAVE_format .
  • ALL:NONE an additional new item to completely switch off time integration of LIQUID including all items like CODI, TURBULENCE and establishing the differential operator stencils. Provides a better performance if stream lines are to be computed after stopping flow simulation in LIQUID context. See TimeIntegration
Tickets:
  • #102884: [Bug fix] no values in Timestepfiles, see here the new option TIMESTEPFILE:ON and TIMESTEPFILE:OFF as a quickfix.
  • #109568: MESHFREE encountered a hard breakdown if reduct() was used before the PointCloudReduction was effected at runtime (can happen for some SAVE_ITEM , that uses reduct() before the first time step executed).
  • #112677: [Bug fix] linked to rigidbody timesteps in MassSpringDashpot.
  • #112748: [Bug fix] EnSight gold readin of pointcloud without timevalues is fixed.
  • #112836: [Bug fix] Multiple, incorrect entries in TimestepFile occurred in case of initially empty point clouds.
  • #112926: [Bug fix] Incorrect initialization of points using BErand().
  • #112974: Prio: corrupted simulation results in (Ensight und ERF) in current main . Problem was due to empty pointcloud in the beginning of the simulation. Fixed for ensight6 binary and ERFH5 .
Bugfixes:

beta2025.06.0

New:
  • A begin_construct_always{ environment can now be defined which automatically switches between begin_construct{ (normal runs) and begin_construct_atRestart{ (restart runs).
  • When optional parameters are specified for %MOVE_MassSpringDashpot%, the total spring and damping forces account for contributions from the spring and damping constants.
  • Avoid the construction of an "inflow pile" (Nb_InflowLayers of points on top of a %IDENT_inflow%-boundary) if the boundary part is not active from the beginning. See also %ACTIVE_interval%.
  • New functionality in approxY(): whatToApproximate=7 returns the number of neighbors in the local point stencil.
  • Enhancements in the NEIGHBOR_FilterMethod approach for BE-based filtering, aimed at improving robustness near sharp or complex geometries.
  • RIGIDBODY: new features to fix up to three translational and up to three rotational degrees of freedom of rigid bodies. See the keywords %RIGIDBODY_FixVelocity% and %RIGIDBODY_FixOmega% .
  • Introduced an optional argument in RIGIDBODY_ForceToApplyOnBody_BE to enable caching of area calculations.
  • Added additional stopwatches to monitor the performance of the RIGIDBODY and its associated force-mappings, see NamesOfStopWatches.
  • SAMG: Upgraded to version 2024.4
  • SAMG: suppressed output. If output is desired via settings in an '.in' file, from now on set "mode_mess=0" in addition to "iout=2" and "idump=1".
Tickets:
  • #105558: Provide new executables for HLRS Vulcan architecture due to shutdown of Hawk. We provide tailored executables for the AMD Epyc Genoa nodes on the HLRS Vulcan machine.
  • #110853: Fixed bug for reading of SCAN_ClustersOfConnectivity. The bug ONLY occurs if SCAN_ClustersOfConnectivity is defined in common_variables. To be safe with the pervious version, users should therefore rather define this parameter in USER_common_variables.
  • #108782: Trying to improve inside-outside check for points, disrupting from an angle-motivated edges, especially in non-unique geometry situations (changing their status from wall to interior, but landing again at a wall). This change does not address disruption at open edges.
  • #108737: Bugfix for volume correction, especially for the VOLUME_correction_FreeSurface functionality, which is based on the correct computation of %ind_BNDfree_defect%. This was, by accident, reset to zero in the presence of several chambers.
  • #111332: Undefined breakdown of the code during startup (read in of wildcard-PIDs). This was perhaps due to a compiler bug that only became apparent now. The critical point was in the code for a long time without causing any problems.
Bugfixes:
  • Fixed memory-related issues for the new version of the NEIGHBOR_FilterMethod, see Version=3+BE.
  • Fixed problem for right-hand side containing a big number of elements, especially when using CODIV functionalities.
  • Fixed wrong initialization of the "last" timing for the automatic solver control in the first time step.
  • weakly compressible materials: there was a bug in the computation of the hydrostatic pressure. The compression module was, by mistake, added twice to the diagonal of the system matrix. This might have led to a slower adaption of the hydrostatic pressure to the gravity.
  • if the user gave a pressure boundary condition BC_p()$...$) = ( %BND_NEUMANN%, A ), then automatically the condition for the correction pressure has to be BCON($...$,%ind_c%) = ( %BND_NEUMANN%, 0 ) , important is the zero. Due to a mistake, the condition was set to A instead of 0. Luckily, most users rather use BC_p=%BND_wall%, in this case, no problem occurs. For older versions, avoid this problem by explicitly defining the boundary condition for the correction pressure by BCON($...$,%ind_c%) = ( %BND_NEUMANN%, 0 ) in the input file.
  • Fixed bug in evaluation of %SUMMATION_BND% integrations.
Known Limitations:

beta2025.03.0

New:
  • A new option has been added to retrieve xCOG and vCOG within the context of %MOVE_MassSpringDashpot%. This enables the calculation of the modified spring length and its rate of change at each RIGIDBODY sub-timestep.
  • PAMCRASH Reader: An option to readin line segments via the keyword BEAM is added. (experimental)
  • Redesign of automatic linear solver control: Fixed several mal-decisions (e.g. while interfering with an outer linear solver test loop); if the Output Level >=3, inform user about WHAT decisions are made and WHY
Bugfixes:
  • If RIGIDBODY_ForceToApplyOnBody_PC and RIGIDBODY_ForceToApplyOnBody_BE are assigned to the same MOVE-flag, RIGIDBODY_ForceToApplyOnBody_BE takes a priority. If no keyword is defined, RIGIDBODY_ForceToApplyOnBody_PC is used by default.
Known Limitations:
  • Changing the options for Automatic Solver Control using Computational Steering is experimental.

beta2025.02.0

New:
  • From now on, the MESHFREE simulation ID set by the common variable iFPM_process_ID can be any string with a maximum of 64 characters.
  • MESHFREE.InputFiles.USER_common_variables.PhysicalProperties.mue.Hyperelastic stress tensor model e.g. for modelling of snow
Tickets:
  • #106346: New option SAVE_skipFirstTria as workaround for ERFHDF5 on HAWK.
  • #109044: Automatic linear solver control: Negative _nn_ forces the current solver for at least |_nn_| time steps, or until the maximal number of failures is reached. See also LINEQN_solver, LINEQN_solver_ScalarSystems.
  • #109044 , part 2: If using SAMG, SPM_Regularization_Type = 2 is reset to 1
  • #109370 Revise approxY(): The user has the option to specify a negative approximation order, which facilitates order reduction in cases where the stencil may yield an unstable approximation.
  • #108859 Fix problem of beta2024.11.0 with %MOVE_MassSpringDashpot%: There was a data mismatch due to recent changes for optional spring force and damping force in %MOVE_MassSpringDashpot%
Bugfixes: Known Limitations:
  • Automatic linear solver control: In the case of negative _nn_ , the selection of the fastest solver at the prescribed interval is not always carried out correctly. See also LINEQN_solver, LINEQN_solver_ScalarSystems.

beta2025.01.0

New:
  • A missing geometry file now causes an error termination of MESHFREE. For backwards compatibility one can set the environment variable 'MESHFREE_IGNORE_MISSING_GEOMETRY="true"' to restore the deprecated behavior.
Tickets:
  • #109044: Rebuild SAMG structures less frequently to avoid potential MPI deadlocks.
Bugfixes:
  • The failing usage of the environment variable MESHFREE_Ucv_DEFAULT is now recovered with MESHFREE_UCV_DEFAULT.

beta2024.11.0

New:
  • Implemented apparent heat capacity method (ApparentHeatCapacityMethod) for inclusion of latent heat in phase change simulations.
  • Optional arguments added for MassSpringDashpot to directly provide a spring and damping force.
  • Force-mapping keyword has been changed for the RIGIDBODY solver. New keyword for RIGIDBODY_pressureToApplyOnBody is RIGIDBODY_ForceToApplyOnBody_PC.
  • A new force-mapping keyword RIGIDBODY_ForceToApplyOnBody_BE is added. The user provides the force components directly. RIGIDBODY_ForceToApplyOnBody_PC takes priority if both are defined for the same MOVE flag.
  • Deprecate COMP_SortBEintoBoxes_Version = 1 for 3D, only applicable for 2D (nue = 2 is no longer fully supported).
  • Deprecate ORGANIZE_ActivateBNDpoints_Version = 2.
  • Deprecate ORGANIZE_DistanceToBoundary_Version = 2.
  • Implemented a new algorithm for the version 3 of the NEIGHBOR_FilterMethod based on BEs. This algorithm filters the invalid neighbors based on the boundary elements.
Tickets:
  • #106304: Allow a linear solver to fail for a prescribed number of TS. Afterwards, all options will be tested
  • #107536: Fix error in BEmap.
Known Limitations:
  • RIGIDBODY_ForceToApplyOnBody_PC is the new keyword for PC-based force-mapping instead of RIGIDBODY_pressureToApplyOnBody, existing setups must be updated with this change.
  • LINEQN_solver_ScalarSystems cannot be set to 'AUTO' for both temperature and TURB in the same simulation.

beta2024.09.1

New:
  • Switched licensing over to FlexNet Publisher (FlexLM). Please check InstallationGuide for instructions and contact Support if you require assistance.

beta2024.09.0

New:
  • Windows version available (experimental).
  • SAVE_BE_filter: speed optimized version.
  • SAMG: More robust handling of singular linear systems
  • timeAvg(): implement the Fourier transform in order to check frequencies of acoustic sound waves.
  • approxY(): setting default value 1.3 for factor_allowed_overshoot, aiming for more stable results. If extrapolation is desired, manually set it to a higher value or 0 (for unlimited).
  • Co-Simulation with LS-DYNA:
    • Force-mapping keywords have been changed for co-simulation with LS-DYNA. New keyword for ForceToApplyOnBody is LSDYNA_ForceToApplyOnBody_BE and for PressureToApplyOnBody is LSDYNA_ForceToApplyOnBody_PC.
    • A different shell thickening value can be provided per alias (or PID) for shell elements. The feature is specific to the interface with the co-simulation tool LSDYNA (experimental).
    • The keyword LSDYNA_ScaleMinNodeDistance is added to merge nodepoints on MPI-borders. This value allows the user to decide the tolerance to be used for merging nodepoints on the interface with the co-simulation tool LS-DYNA.
    • Solid stress and turbulent stress keywords have been changed for co-simulation with LSDYNA. New keyword for SolidStressWeight is LSDYNA_SolidStressWeight and for TurbulentForceWeight is LSDYNA_TurbulentForceWeight.
    • LSDYNA_ForceToApplyOnBody_PC uses the PC index, ind_Sn to map pointcloud data on BEs, if the third optional argument > 0.0.
  • approxY(): setting default value 1.3 for factor_allowed_overshoot , aiming for more stable results. If extrapolation is desired, manually set it to 0.
Tickets:
  • #106330: Fixed deadlock in SAMG
  • #107220: Allow wildcards in BErand()
Bugfixes: Known Limitations: In the co-simulation with LS-DYNA, the keywords for force-mappings have been modified. LSDYNA_ForceToApplyOnBody_BE is the new keyword for BE-based force mapping, whereas LSDYNA_ForceToApplyOnBody_PC is the new keyword for point-cloud based force mapping. Existing setups must be updated with this change.

beta2024.07.0

New:
  • Enhance the functionality of BC_WettingAngle by the definition of a dynamic contribution to the contact angle of free surfaces in contact with walls.
  • bubble pressure can now be given as a boundary condition to the free surface EITHER to the hydrostatic OR to the dynamic pressures. Thus, one might switch off calculation of hydrostatic pressure and being able to only live with the dynmic part. See BubbleCorrectionPressure
  • Random numbers: DEBUG_GeneralParameter(76) is no longer used to switch between PRNG generators.
Tickets:
  • #96537 and #104477: new option ReadToPCStandby for EnSight ReadInPointCloud .
  • #105306 Version beta2024.05.0 contained a bug for RIGIDBODY .
  • #105577 MF will be aborted if the MKS solver makes higher number of rigid-body sub-iterations then the upper-limit specified with the parameters in RIGIDBODY_TimeIntegrationVersion.
  • #105774 (poor performance at startup due to extensive loops including singleton boundary elements (BND_tria &whatever& x1 y1 z1 x2 y2 z2 x3 y3 z3):
    • the loops are skipped during read-in in case they are not needed (such as scanning of aliases does not need the loops in the begin/end_boundary_element{} - environment -> save time for interpretation of non-dedicated lines
    • the BE_decouple-algorithm is called at the end of reading the singleton boundary elements -> saves a lot of computation time
    • DEBUG_GeneralParameter(94)=0 will switch this improvement off (just for performance-reference)

beta2024.05.0

New: Tickets:
  • #96537: Improper access to an internal array prevented.
  • #103974: The implementation of the math function atan2() was corrected and verified.
  • #104477: A first memory optimization when reading in a STANDBY Pointcloud.
Bugfixes:
  • Big TIMECHECK output files were generated when MF-LSDYNA coupling was used.
  • Timecheck for the LSDYNA.Extern produced incorrect results when used with the CouplingRate.
  • Introduction: fix hyperlinks in linked documents, fix standard solution of Task 19 of the Training Setup.
  • RIGIDBODY_TimeIntegrationVersion: bug in the reading of the parameter list: if not all optional parameters were given, the defaults for the optional parameters were overwritten.

beta2024.03.1

Bugfixes:
  • BEmonApprox(): this function was no longer properly called due to an incorrect warning for unknown acronyms.

beta2024.03.0

New: Tickets:
  • #86046: the two digit-mode of approximation order in approxY() was verified.
Bugfixes:
  • BDF2 for velocity ( V:BDF2 ): did not work safely before due to wrong allocation of BDF2-indices storing the velocity values of previous time steps
  • Compatibility between LSDYNA and USER_h_PC_refinement
  • for the heat production due to viscous friction, ONLY use the laminar viscosity %ind_ETA% (resulting in %ind_diss%). Before, %ind_ETA_eff% was used wrongly, taking into account also the turbulent partition of the viscosity. Maybe a minor, problem, but just for correctness.
  • The default of the second component of COMP_SkipHighVelocities was not always properly set.
  • add the heatsource term \( q\) to the %BND_ROBIN% boundary conditions, which was missing there in the formulation. Problems did not appear previously, if the control thickness was set to zero (default). Problems did also no appear, if the neat source was zero (which is the case in most applications).

beta2024.02.0

New: Tickets:
  • #99356: Activation of boundary points on moving geometry parts for EULERIMPL fixed.
  • #102484: reading-in of very big restart files: in the past, having many MPI-processes together with many saved MONITORPOINTS , the reading of a restart file could have failed. Capturing round-offs in integer divisions now fixes the problem.
  • #102082: now, initialize undefined acronyms ($...$-variables) with a negative index. Thus, avoid consistency problems in case of very many acronym definitions (>400) in the input file
Bugfixes:
  • Revert default to GEOTREE2_EstablishCON_Version = ( 3, 3 ) to fix problems in some stable tests. This was last changed in beta2023.08.0.
  • bugfix of a hidden bug, that rarely appeared: InDom( "Alias" ) could have led to segmentation faults, if the index of "Alias" was bigger than 2500.
  • Fixed bug for %MONITORPOINTS_CREATION_ByCluster% that could lead to creation of cluster monitor points even though the creation condition is not satisfied for all points of the cluster.
  • Fixed bug in DROPLETPHASE_LockedPoints_Collision. There were no locked points created.

beta2024.01.0

New:

beta2023.12.1

Bugfixes:

beta2023.12.0

Known Limitations:

  • Experimental case insensitive reading was enabled for Ucv reading. Unfortunately, we observed negative side effects and do not recommend using this version.

New:
  • Additional option in DropletSource to apply a RHS expression such as propability density functions for the initial velocities
  • The cv variable 'STRESSTENSOR_NumberSubcycles' is deprecated.
Tickets:
  • #101715: events triggering restart, stop, abort, resume are no longer effective during the PC refinement phase.
  • #101594: Limit TurbLengthScale Not Working. If a negative sign in DEBUG_GeneralParameter(54) was used (such as DEBUG_GeneralParameter(54)=-1.5E-7), then MESHFREE steps into a specialized KEPSILON-repair function. Here, the restriction of the turbulent length scale (see COMP_development(1) digits 6...8) was not yet implemented. This now is fixed.
Bugfixes:
  • COMP_DoOrganizeOnlyAfterHowManyCycles: parameter was not correctly overwritten in USER_h_PC_refinement refinement phase.

beta2023.11.0

New: Tickets:
  • #101264: Representative mass loss at chamfers.
  • #101268: bug aero kepsilon linear solver does not converge: the current changes do not strictly avoid the problems of ticket #101268, however some improvements have been made. again: a physical limit/restriction for the turbulent layer thickness %ind_turbulentWallLayer% seems to be given, see section 2.2. and especially equation (2.12) in the document DOCUMATH_NumericalIntegrationOfTurbulence.pdf
  • #101308: Deprecate COMP_DoOrganizeOnlyAfterHowManyCycles_STANDBY and automatically skip organization steps for the STANDBY point cloud if possible (ensuring compatibility).
  • #101352: timeAvg() did not work in the 4-Parameter-version (beta v23.10.0).
  • #101369: Reproducible Random number not properly set. Use DEBUG_GeneralParameter(76) = 0.0 to use old version of random number generation.
  • #101428: SAVE_format BLENDER now supports ONLY:PARTICLES, GEOMETRY:OFF and GEOMETRY:ONCE options
Bugfixes:
  • added CV variable SAVE_ERF_retryWrite. If set to 1, MESHFREE will continuously try to write the ERF file even if another program is using the result file instead of crashing.
  • There was a leak of representative masses %ind_mi_rep% (see RepresentativeMassAlgorithm) in the case when points were removed due to wall distance, but have been isolated at the same time (Y%ind_nbRegularNeighbors%=1). See also ticket #101264.
  • ticket #101352 was clearly a bug, that is now fixed. In beta2023.10.0 , the 4-parameter-version ( timeAvg( t, u(t), index, Dt ) ) did not work. Only the 5-parameter-version worked ( timeAvg( t, u(t), index, t-dt, t) ) and can be used as workaround.
  • Interoperability of COMP_DoOrganizeOnlyAfterHowManyCycles and USER_h_PC_refinement: pointcloud organization was previously also skipped in the refinement phase.
  • Bugfix in SAVE_format BLENDER, when writing out boundary elements

beta2023.10.0

New:
  • ord_convection: reimplemented option for EULER. Enables users to go back to the linear-order-implementations of versions before 2022, in order to compare solutions.
  • Trigger option for CODI_eq (third parameter, optional), to better handle expensive evaluations.
  • Trigger option CODI_executeNow in general for CODI definitions.
  • smoothing of the representative density DefinitionRepresentativeDensity . Avoid buckling of free surfaces due to VOLUME_correction_local .
  • see DEBUG_GeneralParameter(82) for controlling the computation of Y%ind_Sn(1)% ... Y%ind_Sn(3)%
  • Upgraded SAMG to Release 2022.8
  • new function BErand() for generating point samples on boundary parts.
  • TurbulenceModels: change the default value of DEBUG_GeneralParameter(57) = 2 (previously 1) -> that pushes k and epsilon in such a way, that the turbulent viscosity will remain nearly constant in the following time cycle.
  • Feature USER_h_PC_refinement has been refined. (turning off rigid body solver, timecheck statistic writeout)
  • COMP_development(1)=91001010 (default) providing now also the option to cut the turbulent length scale with respect to the local SMOOTH_LENGTH
  • Add the four-quadrant inverse tangens atan2() as mathematical function to be used in user defined equations.
  • timeAvg() - new function in MF_Function to compute time averages of scalar properties
  • LINEQN_BundleSolvers - possibility to bundle linear solver items for the persomance analysis in the context of LINEQN_solver_ScalarSystems = ... AUTO: ...
  • SAVE_BE_NODE_ITEM - interoperability with begin_save{
Tickets:
  • #97289: Dropletsource for boundary elements, see docu of BErand() for application.
  • #98663: Save_results when velocity solver does not converge: additional option in COMP_SkipHighVelocities.
  • #100136: exploding value for %ind_tauW% in case of negative/huge values of %ind_TurbulentWallLayer% . %ind_TurbulentWallLayer% is now truncated to the interval (0,1] .
  • #100647: values not written to timestep file before the 10th timestep.
  • #100793: confusing/wrong output/behavior for k-epsilon concerning the linear solver automatism, see LINEQN_solver_ScalarSystems
  • #100796: Enhancement of solver automatism for linear systems of (hydrostatic, correction, dynamic) pressure, see LINEQN_BundleSolvers
  • #101264: rep. mass loss at chamfers
  • #100819: inconsistent computation of representative masses at %IDENT_inflow% , see BugFixes
Bugfixes:
  • For *.msh files exported from Gmsh in format Version 4.1 (ASCII) not all nodes were properly read.
  • inconsistent computation of representative masses at %IDENT_inflow% boundaries fixed. Now also take into account the points, which carry Y%ind_kob% = %BND_edge%
  • BEmonApprox(): changing the caching strategy for better interoperability with ERFHDF5. We have observed performance issues when using multiple different BEmonApprox()-statements with ERFHDF5.
  • Fix SAVE_filter for BLENDER SAVE_format.
  • MPI Shared Memory: Increase maximum number of boundary elements.
  • There was a leak of representative masses %ind_mi_rep% (see RepresentativeMassAlgorithm ) in the case when points were removed due to wall distance, but have been isolated at the same time (Y%ind_nbRegularNeighbors%=1). See also ticket #101264

beta2023.08.0

New: Tickets:
  • #98788: Revision of AbaqusInterpolation (feature is not maintained) documentation, fix problems during writing of interpolation output.
  • #99205: Prevent SAMG from returning Not-a-Number if a singular filesystem is encountered in a small (<~100 points) decoupled subsystem.
Bugfixes:

beta2023.07.0

New: Tickets:
  • #96615: Remove upper bound for number of arguments in equations.
  • #99123: ReadInPointcloud: Stationary fields could not be loaded in beta2023.05.0 (if you require a workaround contact support.)
  • #96909: DROPLETPHASE solver LiquidLayer: Use analytical integration for the viscous forces (decceleration) in normal direction to the boundary layer
Bugfixes:
  • BEmon_MPIcomm_Version = 2 or 23 does not yield results when shared memory is turned off by COMP_SharedMemoryForBE = false. In this case, we switch now to the old default BEmon_MPIcomm_Version = 1.
  • CUMU_AVERAGE delivered wrong values, when the boundary elements were initially hit by multiple monitor points.
  • CUMU_MINIMUM and CUMU_MAXIMUM delivered wrong values due to a wrong initialization.

beta2023.05.0

New:
  • Output Level: New EnvironmentVariables MESHFREE_OUTPUT_LEVEL and CommandLine option --output-level, -ol. Assigning all existing printouts explicit levels is work in progress.
  • Deprecated N_addvar and usage of %ind_addvar()%. Use UserDefinedIndices instead.
  • Changed log folder name from 'FPM_log_ ...' to 'log___SIM_ID_'.
  • Changed value of internal variable %IDENT_locked% to 1.
  • New cumulation features for SAVE_BE_MONITOR_ITEM: %CUMU_MAXIMUM%, %CUMU_MINIMUM%, %CUMU_AVERAGE%.
  • SAVE_BE_MONITOR_ITEM, %CUMU_INTERVAL%: reset the cumulation interval by formula, keyword %CUMU_USER_INTERVAL%.
  • New cv Parameter BEmon_DefaultValue to manually set the reset value (experimental).
  • New option for cv Parameter BEmon_MPIcomm_Version = 3 (experimental).
  • Default value of BEmon_MPIcomm_Version changes from 1 to 2.
  • Default value of BEmap_MPIcomm_Version changes from 2 to 3.
  • Adaptive substepping algorithm for LSDYNA and MESHFREE coupling (experimental).
Tickets:
  • #95001: restart_toberemoved can now also be used with wildcards.
  • #95981: Fix determination of consistent surfaces (%ind_dA%) and allocation for boundary filling.
  • #96443: Reactivate ADMIN_TIME_INTEG.SAVE.UDPmonPrep, see NamesOfStopWatches.
  • #97953: Clarify the documentation of ReadInPointCloud. Only data from EnSight files can be used for STANDBY point clouds, whereas data from ASCII files trigger a point cloud quality check and subsequent termination of the simulation.
  • #98063: Introduce COMP_DoOrganizeOnlyAfterHowManyCycles_STANDBY (experimental) to enable skipping of STANDBY point cloud organization.
  • #97468: Fixed Bugs in DROPLETPHASE_tools: Locked Points were not detaching properly.
Bugfixes:
  • SAVE_format: TIMEACC now allows reading two digit numbers (up to a maximum of 92).

beta2023.03.0

New: Tickets:
  • #91750: Fixed unclean MPI exit.
  • #96042: Add option to define IDENT flags for alias definitions as a function of e.g. time, boundary condition flag, midpoint of boundary element (experimental).
  • #96064: BubbleAlgorithm: Keep irregular bubbles for a specified number of time steps with nb_TStoRescueInvalidBubbles
  • #96550: SAMG: Fixed treatment of small components where the pressure is only determined up to a constant.
  • #96652: OBJ Reader will by default split quads into trias, and the normal orientation is flipped.
  • #96790: In beta2023.02.0, the InDom() function did not work properly. This has been fixed now.
Known Limitations: The orientation of geometry parts read in from OBJ files has been reverted compared to previous versions. It is now consistent with other geometry file formats. For existing setups, please fix the orientation with revOrient{}. To check any geometry orientation, use CONTROL_StopAfterReadingGeometry with SAVE_format code 'NN--' or SimCut.

beta2023.02.0

New: Tickets:
  • #96165: there were two bugs related to multisave ERF causing aborts, both now fixed. Varkeys for the triangulation were duplicated and a corrupt memory in combination with SAVE_BE_filter occured.
Bugfixes:
  • Fix license check with CommandLine option --check-license, when running with multiple MPI processes.
  • LSDYNA: Fix license check for floating license model when running with multiple MPI processes.

beta2023.01.1

New:
  • Allow user provided point weights for MPI bisection (domain decomposition). See MPI_BisectionWeight .

beta2023.01.0

New:
  • DROPLETPHASE: first complete version of locked points for snow-cluster modelling, see DROPLETPHASE_tools. Especially see the convenience function for a simple DEM-implementation in BuiltIn_DEM_method.
  • Updated compilation procedure to allow support for more operating systems.
Bugfixes:
  • Multiple smaller bug fixes.
List of members:
All Complete release notes for the MESHFREE beta executables
© 2015-2026 Fraunhofer ITWM and Fraunhofer SCAI
Index
Publishing Notes
Data Protection
Index
Publishing Notes
Data Protection
© 2015-2026 Fraunhofer ITWM and Fraunhofer SCAI