iconCFD MESH is an automatic hex-dominant mesher delivered by ICON as part of the CFD service iconCFD. It is based upon open source CFD technology and is licensed under GPL. This document summarises the modifications, improvements and known issues for this release.
- New improved mechanism of load balancing during parallel execution of all components of the code. Parameter maxLoadUnbalance (default value 0.2) exclude expensive domain re-balancing stage if overall value of load unbalance is less than the given value. Provides estimated 10-15% speed increase during parallel execution of iconCFD Mesh due to omission of re-balancing where possible.
- New improved triangulated surfaces handling procedure has been developed and implemented, providing approximately 10% speed improvement in meshing stage.
- New option introduced in iconCFD Mesh to control the first cell layer thickness by patch.
- New utility checkVolMesh performs analysis of the volumetric cell shapes by computation of the L2 norm of the iso-parametric transformation form parametric (ideal cell shape) to physical space (deformed cell shape).
- New version of the mesher supports scotch final domain decomposition.
- New option to redistribute the mesh at the end of the generation by adding an option to the meshDict - doRedistribute - set to true or false. If true, the code will look for an extra dictionary - decomposeParDictFinal - which is same by format to decomposeParDict and defines final domain decomposition (i.e. scotch, metis). Recommended workflow is as follows: iconCFD Mesh with hierarchical decomposition (M 2 1), where M is (total number of domains)/2, then final domain decomposition M*2 with scotch. Final redecomposition is parallel and DOES NOT require reconstruction of mesh on one node. If doRedistribute is set to false - the mesher will behave in the same way as previously.
- parMapFields is completely re-coded, eliminating multiple re-loading of source domains from file and improving speed of operation up to 30%. No changes in the way it is used.
- Proximity-based refinement implemented. Controlled by specifying number of additional refinement levels over and above the standard surface max refinement level that may be applied to try to prevent multiple parts of the geometry passing through the same Cartesian cell.
- Enhancements to feature line snapping algorithms.