A spectral-geometry approach to detecting structural mismatch in time-series systems.

SOHM Research Labs develops a spectral-geometry diagnostic basis for comparing time-series signals beyond power spectral density and autocorrelation.

The method extracts four coordinates from delay-embedded trajectory matrices — Gap, SMK, Alignment, and Entropy — to measure structural differences in signal dynamics.

In controlled experiments on turbulence data, different error types produce distinct signatures across the basis, enabling detection and classification of structural mismatch.

Primary Applications:

- Digital twin validation

- Simulation model comparison

- Solver regression testing

- Process monitoring

What This Enables:

- Detection of structural mismatch missed by PSD-based metrics

- Consistent signatures across distortion types

- Localized identification of model failure

- Basis-level comparison across simulations

Evaluation:

A technical demonstration package is available upon request, including:

- Runnable comparison script

- Example signals

- Visual diagnostics

- Reproducible test cases