Design of domain shapes

A few design goals govern how MeshFreeFoam shapes are implemented, with the most important ones being:

• Support for dynamically loaded models, starting from STL files
• Support for basic geometric operations (adding, subtracting, scaling, rotating, …, etc.)
• Support for physics on boundary patches as well as easy MPI comms
• Customizable boundary filler, supporting both single-layer and multi-layer filling
• Customizable rho-based inner fillers.

Also, shape classes store no actual grid data to facilitate geometric operations. Dedicated mesh classes will store point coordinates hierarchically.

graph LR
    grid -- as baseShape --> region1
    grid -- as baseShape --> region2
    region1 --> boundaries1
    region1 --> interface
    region2 --> interface
    region2 --> boundaries2
    boundaries1 -.- grid
    boundaries2 -.- grid
    interface -.- grid
    innerFiller o--o grid
    boundaries1 & boundaries2 & interface o--o boundaryFiller

Boundary Treatment

Initially, only single-layer boundary layers are implemented. Shapes only read the boundary’s interface type, which should be similar to OpenFOAM boundary patches.

At the time of writing, a few boundary types are planned:

• fixedWall for stationary walls.
• elasticWall for “elastic” surfaces.
• freeSurface for free surface interfaces.
• mixedBoundary for mixed Neumann-Dirichlet boundary conditions.
• genericBoundary for solving arbitrary transport PDEs on the boundary. PDEs are aggregated for later implicit solving with the whole system.

Here is a list of requirements on boundary patches:

1. Awareness of how many sides (1 or 2). The second side can be imaginary if at the processor’s boundary.
2. Access to the equation operator classes; however, this may result in circular dependencies.
3. Access to neighboring inner grid points and data.

• ←Previous
• Next→