Can a direct global linear basis plus simple phase/speed interpolation provide a transparent, deployable baseline?
PARAMETER STUDY A · FIXED AoA = 12°
U Sweep
Seven transient CFD anchors from 50 to 200 km/h provide a controlled test bed for spatial compression, coherent/residual decomposition, spectral modes and autonomous reduced dynamics.
7 speeds4,480 snapshots96,000 cells4 model/dynamics tests
ENGINEERING QUESTION
Which representation is useful—not merely compact?
The U Sweep deliberately keeps geometry and angle of attack fixed so model behavior can be compared without changing the underlying physics configuration. Each method asks a different question of the same CFD data.
Does explicitly separating repeatable coherent motion from secondary residual dynamics improve structure or deployment?
Does conditioning the decomposition on frequency reveal a much simpler coherent representation?
Can POD coordinates form a closed autonomous dynamical state rather than relying on prescribed phase?
METHODS IN PARALLEL
Different outcomes are part of the result
PRODUCTIVE · REFERENCE
Raw POD
99.5% POD ranks: u 37, w 56, Cp 16, ω 28. Surface loads remain extremely compact. Current live demonstrator is deployed.
Open project + demonstrator →PRODUCTIVE · CHALLENGER
Phase + Residual POD
Phase alignment substantially improves repeatability. The coherent field becomes more compact, while the residual reveals richer broadband/secondary structure.
Open project + demonstrator →PROMISING · DIAGNOSTIC
SPOD
The fundamental frequency is strongly rank-1 dominated and follows a consistent convective scaling across speeds. A full cross-speed generator remains future work.
Open SPOD result →MIXED · PARKED
SINDy
A sparse rank-2 U125 velocity oscillator works; higher-rank and surface autonomous systems do not reliably close. The state architecture is the key unresolved issue.
Open SINDy result →CROSS-METHOD TAKEAWAY
The hard part is increasingly temporal/state structure—not only spatial compression
POD works well
All tested approaches confirm that important field content can be represented in a reduced spatial basis, especially coherent velocity and surface quantities.
Strong periodic family exists
Cl/Cd phase diagnostics and SPOD both identify a consistent dominant aerodynamic oscillation across the seven speeds.
Not just one extra harmonic
The phase-residual analysis shows structured 2–6 f0 content together with substantial broadband/irregular residual variance.
Closure matters
SINDy shows that a low-dimensional oscillator can be discovered, but higher fidelity requires choosing a state with sufficient closure and possibly memory/coupling.
NEXT PARAMETER LAYERS