Picid: Modular Evaluation Infrastructure for Reproducible PHM
Picid, a novel modular evaluation framework, seeks to enhance reproducibility and standardization in Prognostics and Health Management (PHM) research. This system articulates the PHM evaluation process as a clear, executable, and reproducible protocol, tackling challenges like implicit data divisions, preprocessing, label synchronization, temporal windowing, and performance metrics. Picid guarantees the construction of deterministic, leakage-safe datasets while offering adaptability across various PHM applications, such as fault detection, diagnostics, and prognostics. It allows for the integration of new datasets and model types without breaching protocol invariants, standardizing evaluation boundaries and data contracts. This framework has been detailed in a paper on arXiv (2605.28345) to facilitate better comparability and reproducibility of findings.
Key facts
- Picid is a modular evaluation infrastructure for PHM.
- It formalizes the PHM evaluation pipeline as an explicit, executable, and reproducible protocol.
- It enforces deterministic, leakage-safe dataset construction.
- It supports fault detection, diagnostics, and prognostics through a unified interface.
- It can be extended to new datasets and model classes without violating protocol invariants.
- The framework addresses implicit protocol choices like data splits, preprocessing, label alignment, temporal windowing, and metrics.
- The paper is available on arXiv with identifier 2605.28345.
- Picid aims to improve reproducibility and comparability in PHM research.
Entities
Institutions
- arXiv