How it works¶

Overview¶

InterpCore maps values known on a source point cloud (e.g. the integration points of an electromagnetic or CFD model) onto a destination point set (the nodes or the element centroids of a mechanical FE model). The neighbour query is computed only once, in the constructor, since all source files are assumed to share the same coordinates.

Two mapping philosophies¶

Kernels fall into two families, depending on which point set the KDTree is built on and on which point set the loop runs.

Orange circles are source points, blue squares are destination points.

Forces are extensive quantities: the sum of the nodal forces must equal the total force of the source, so each source force is split among the destination nodes. Heat flux, heat generation and HTC are intensive quantities (densities): each destination point must receive a representative local value, regardless of how many source points are nearby.

Per-point decision flow¶

For every point of the loop, the following logic is applied (see interpolate_block):

flowchart TD
    A[Point of the loop] --> B{Neighbours found<br/>by the query?}
    B -- no --> N
    B -- yes --> C[Discard neighbours farther<br/>than max_distance]
    C --> D{Any neighbour left?}
    D -- no --> N{Philosophy}
    N -- source-to-target --> U[Add value to<br/>unmapped total]
    N -- target-to-source --> E{accept_no_neighbor?}
    E -- yes --> Z[Assign 0]
    E -- no --> R[[Raise InterpolationError]]
    D -- yes --> F{"Closest neighbour within<br/>coincidence_tolerance?"}
    F -- yes --> G[Copy the value directly]
    F -- no --> H[Apply the kernel]