ME5311 Project 1

This project is a compact Python workflow for vector-field analysis, covering POD, spatial spectrum analysis, temporal PSD analysis, and publication-style figure export. I built an end-to-end and reproducible pipeline from raw snapshots to final outputs, including data loading with validation and subsampling, mean/fluctuation decomposition, POD via SVD, low-order reconstruction checks, isotropic and 2D spectral analysis, dominant-frequency extraction, anisotropy quantification, and automatic export of key metrics.

The main results show that dominant structures are captured efficiently by POD, with 90% and 95% cumulative energy reached at modes 12 and 32, and a 4-mode reconstruction relative error of 0.2832. The spatial spectrum indicates a clear dominant scale at |k*| = 0.0625 with characteristic wavelength 16, segment-based spectral peaks remain stable, anisotropy is pronounced with Ex/Ey = 5.79983, and leading modal dynamics are concentrated in low frequencies around 3.33e-4 to 2.67e-3.

Project Files

• main.py: full analysis pipeline entry point
• config.py: paths and analysis parameters
• load_data.py: data loading and frame indexing
• analysis.py: POD, spectra, PSD, and metrics functions
• plot.py: consolidated publication-style figure generation
• data/vector_64.npy: input dataset (not included — see data/placeholder.txt)

Requirements

Install dependencies:

pip install -r requirements.txt

Run

From the project root:

python main.py

Outputs

• Figure: outputs/q1_q4_summary_B.png
• Text summary: outputs/analysis_summary.txt