Davidson Frame

Abstract

The Davidson Frame—formally the Structured Information Deterministic Manifolds (SIDM) architecture—is a seven‑stage inference framework for reconstructing latent system states under noisy, irregular, and information‑constrained observation regimes. This study provides an initial component‑level empirical validation of the Frame’s admissibility and gating logic using the NASA CMAPSS turbofan degradation benchmark (FD001–FD004). The objective is not competitive Remaining Useful Life (RUL) prediction, but a focused test of whether the Frame’s entropy‑gated admissibility mechanism behaves consistently when exposed to real multivariate degradation data. A deliberately minimal linear Health‑Index model was used for Stage‑5 latent‑state reconstruction, while Stage‑6 applied an entropy‑based admissibility gate that attenuates or rejects candidate estimates when observational irregularity exceeds a calibrated threshold. Forty test engines (ten per CMAPSS subset) were evaluated. The entropy gate activated consistently across all datasets, with cut‑offs in 9/10 FD001 engines, 7/10 FD002, 5/10 FD003, and 7/10 FD004. Across subsets, higher observational entropy produced systematically more conservative latent‑state estimates. Absolute RUL errors remained substantial (mean absolute errors 53.7–107.2 cycles), reflecting the limitations of the intentionally simple Stage‑5 reconstruction rather than deficiencies in the gating architecture itself. The experiment therefore isolates a narrower question: does the Davidson Frame’s entropy‑gated admissibility mechanism operate coherently and predictably under real degradation conditions? The observed behaviour supports that hypothesis while identifying reconstruction quality, threshold calibration, larger‑sample evaluation, and gated–versus–ungated comparison as the next validation steps. The longer‑term purpose of this validation is architectural. The Davidson Frame is designed as a general state‑estimation and information‑admissibility layer within the broader Pathfinder system, including autonomous relay and beacon subsystems associated with the Phase‑Managed Handover System (PMHS) and Poloidal‑Shielded Analog Relay Buffer (PSARB). Within that context, the present study constitutes one independent validation step toward a larger architecture in which sensing, reconstruction, uncertainty assessment, data handling, and autonomous operational decisions are separately testable prior to integration.