NACA0012 Aerodynamics
A layered engineering portfolio: start from accepted OpenFOAM CFD, then build parameter studies, compare reduced-order and data-driven representations, validate what works, and keep failed or inconclusive experiments visible.
One controlled source before model comparison
The current work starts from the same accepted 2-D transient URANS campaign and canonical HDF5 preprocessing. Model tests should differ in representation or dynamics—not quietly in their source data.
OpenFOAM → canonical HDF5
Geometry, volume fields, wall fields, time/phase metadata and exact physical weights used by the downstream ROM experiments.
Seven speeds, four representation / dynamics experiments
AoA is fixed at 12°. Seven CFD anchors cover 50–200 km/h. The methods below are shown in parallel, but their outcomes are deliberately not presented as equally successful.
Raw POD
Direct POD of unaveraged CFD snapshots. Compact, transparent and already deployed as a continuous-speed live demonstrator.
Live demonstrator availablePhase + Residual POD
Cl-aligned coherent phase field plus a richer residual representation and deterministic multi-frequency residual model.
Live demonstrator availableSPOD
Frequency-resolved coherent structures reveal an exceptionally rank-1-dominated fundamental across most fields and speeds.
Generator / LOSO extension not yet completedSINDy
A sparse rank-2 U125 velocity oscillator works well, while higher-dimensional and autonomous surface models expose state-closure limits.
Useful finding, not current deployment pathMove from one-dimensional speed variation to harder physics regimes
The workflow is broader than one airfoil
Potential next directions include ground-vehicle aerodynamics, morphing geometry, and low-Reynolds-number flapping-wing aerodynamics—each stressing a different combination of geometry, unsteady physics and model generalization.