surface_dual_areas/README.md

HiLiftAeroML native-point dual surface areas

This directory describes the case-local boundary_dual_area_<case>.npy payloads used for area-weighted scoring on the native HiLiftAeroML boundary meshes.

Payload contract

For a case named geo_LHCiii_AoA_j, the payload is stored beside its source boundary archive/VTU:

geo_LHCiii_AoA_j/
  boundary_geo_LHCiii_AoA_j.vtu.tgz
  boundary_dual_area_geo_LHCiii_AoA_j.npy

Each payload is a one-dimensional little-endian float32 NPY with one value per native boundary point, in the exact point order of boundary_<case>.vtu. Values have units of square inches (in^2), matching the native coordinate units.

Definition

Each ordered native polygon [v0, v1, ..., v{k-1}] is fan-triangulated as (v0, vj, v{j+1}) for j=1,...,k-2. For each fan triangle t with area A_t, A_t/3 is added to each of its three vertices. Thus

dual_area[i] = sum(A_t / 3 for fan triangles t incident on point i).

Coordinates are promoted to float64 before subtraction and cross products. Accumulation and validation use float64. Only the final payload is cast to float32. Consumers should likewise accumulate weighted metric statistics in float64.

Validation and provenance

cases/<case>.json binds each payload to canonical hashes of the source point, connectivity, offset, and cell-type arrays and records topology histograms, float64 surface-area closure, float32 storage closure, orphan-point counts, payload SHA256, and runtime provenance. manifest.json aggregates completed case records.

Do not reorder the NPY or apply it to an STL, resampled mesh, triangulation with different point order, or processed PhysicsNeMo PDMsh. It is defined only for the corresponding raw boundary VTU.