Decoupling structural regimes from performance degradation in PEM fuel cells using time-lag analysis

Abstract We analyze time‑domain voltage responses to current‑step events from a long‑duration PEM fuel cell dataset and extract a simple operational time‑lag metric, τ63, defined as the time to reach 63% of the step‑to‑step voltage change (amplitude‑normalized). We then test whether a regime transition inferred from τ63 (“τ‑knee”) remains stable across aging/operation epochs, and how it compares to classical polarization‑curve knees that drift with degradation and protocol history. Using a consistent, fully auditable pipeline, we report (i) a persistent two‑plateau τ63 structure versus current with a detectable transition, (ii) a comparatively stable τ-knee current within the analyzed operating window, and (iii) a stronger drift of polarization‑derived knees. The goal is regime separation and auditability rather than mechanistic over‑interpretation. This repository contains the preprint of Paper 1 together with all supporting analysis artifacts. Included are:- the manuscript in PDF (canonical reading version) and Markdown (source version),- the full deterministic analysis pipeline (version v5.4),- derived and preprocessed datasets produced during data cleaning and normalization,- summary tables and figures used in the paper,- configuration, provenance, and integrity metadata. The work is methodological and observational in scope. No mechanistic or technological claims are made. The original raw experimental dataset is not redistributed due to licensing constraints. All included data files correspond to derived analytical outputs. Instructions for full reproducibility using the original dataset are provided. This release is frozen and archived for citation and auditability.