PROJECT 01 · BRANCH B · PLACEHOLDER

NACA0012 · AoA Sweep

A planned one-parameter branch focused on how changing angle of attack alters flow regimes, model assumptions and generalization behavior.

PYTHON REFERENCE IMPLEMENTATION
/projects/naca0012/aoa-sweep
BRANCH B · PURPOSE

Use angle of attack to deliberately challenge the dynamics assumption

The AoA branch should not merely reproduce the speed-sweep workflow with a different x-axis. Changing α can alter separation, shedding structure, periodicity and the dimensionality of the flow, so dynamics classification becomes an explicit gate before model choice.

DYNAMICS GATE

Periodic or not?

PSD, autocorrelation, repeatability and SPOD/DMD-style diagnostics where useful. Do not assume one phase coordinate remains valid.

VALIDATION

Leave-one-AoA-out

Lock withheld α conditions before model selection. Separate interior interpolation from boundary extrapolation just as in the speed branch.

FILL-IN ROADMAP

What this page will eventually contain

2Characterize flow regimes and dynamicsplots / spectra / cycle evidence
3Train POD baseline appropriate to observed regimesphase model only where justified
4Run leave-one-AoA-out benchmarkfields + Cd/Cl + oracle gap
5Benchmark PhysicsNeMo on identical holdoutssame evidence contract
6Add trust / OOD view for α queriesonly after calibration
PHYSICSNEMO POSITION

Second application of the same challenger track

Speed Sweep establishes whether PhysicsNeMo can improve the known POD/coefficient weaknesses on a controlled one-dimensional problem. AoA Sweep then tests whether the learned representation remains useful when the underlying dynamics themselves change.

See cross-branch PhysicsNeMo roadmap →
← Back to current Speed Sweep case study