Altair ultraFluidX 2020 Release Notes
New Features
- New wall modeling framework
- Offering a new shear-based wall modeling framework with a two-way coupled interaction of the LBM solver and algebraic and ODE-based wall models, with and without consideration of pressure gradient effects.
- Rotating overset grids
- Introducing a first version of overset grids for the simulation of truly rotating geometries. The feature can be applied for rotationally symmetric overset grid regions that rotate around one of the principal axes.
Enhancements
- User Interface
-
- Introduction of passive parts
- Providing a new option to define passive parts that will not be considered in the calculation of aerodynamic coefficients and in the calculation of the projected front surface. The sectional coefficients output is also adjusted accordingly.
- Allow reading of multiple STL files
- Providing the option to read multiple input STL files via the input solver deck.
- Modeling
-
- Part-wise definition of wall model intensity for one-way coupled wall models
- When using a one-way coupled version of the ultraFluidX wall model, you can now specify a part-wise intensity parameter to fine-tune the wall model behavior.
- Output
-
- Enhanced aerodynamic coefficients output
- The aerodynamic coefficients file has been updated to a new layout with additional output variables, including separate pressure and friction contributions and the pitch, roll and yaw moments. The file header also now contains additional information.
- Meshing
-
- New threshold for voxels that are very close to the surface
- Below a certain minimum wall distance, voxels that are very close to the surface are treated as solid. The threshold value has been optimized to improve the mesh quality.
- Other
-
- Improved default MCA parameters for Open MPI
- Update of the openmpi-mca-params.conf file with improved values.
Resolved Issues
- Ill-conditioned Smagorinsky turbulence model implementation for very high refinement levels.
- Boundary nodes that were not touching the surface might have had a zero normal vector.
- Segmentation fault when using too many probes.
- Incorrect calculation of sub-grid distances for several specific cases.
- Periodic boundary condition sets no-slip boundaries as default at the intersection with the object (if any).
- Some probes were not connected to the correct voxels.
- No triangle splitting for inverse baffle parts.
- Incorrect time scale calculation in the exponential averaging.
- Sectional coefficient output did not match the object length in case of object-domain intersections.
- Potentially inaccurate normal vectors for under-resolved simulations of non-smooth geometries.
- Potential division by zero in turbulence generator.