# Asset-integrity demo — data sources & provenance

Every asset-integrity figure the MCP returns traces to one of these real, permissively-licensed
public datasets. Raw datasets are **not** committed — only curated, verified slices under
`public/data/asset-integrity/`. Each entry records the source, license, retrieval date, and
exactly which records were taken so a reviewer can re-fetch and confirm.

> Representative public data — not connected to any live plant. Historian/DHS ingestion
> (OSIsoft/AVEVA PI, GE Vernova) is a per-deployment step. Field investigation and a certified
> fitness-for-service assessment are required before any engineering decision.

## Hydrogen embrittlement (Q2) — `hydrogen-hee.json`

- **Source:** *Hydrogen Embrittlement*, NASA/TM-2016-218602, Jonathan A. Lee, NASA Marshall Space Flight Center (April 2016).
- **URL:** https://ntrs.nasa.gov/api/citations/20160005654/downloads/20160005654.pdf
- **License:** US Government work — **public domain**.
- **Retrieved:** 2026-07-01.
- **Slice taken:** the HEE Index (NTS ratio) + qualitative severity for 48 metals from Tables 1–4 (24 °C, 34.5–68.9 MPa H₂), spanning line-pipe steels (X52/X60/X65/X70), pressure-vessel steels (A516/A517-F/A533B), Cr-Mo (4140/4340), austenitic stainless (304L/316/321/347/A286), martensitic steels, Ni-based superalloys (Inconel 718/625/X-750, Waspaloy, JBK-75), K-Monel, and Ti-6Al-4V. Classification thresholds are the report's Table 1.
- **Note:** `heeIndexNts` = notched-tensile-strength ratio (H₂ / inert). `severity` is the report's own qualitative rating (it can weigh RA and EL ratios too, so it may differ from a strict NTS-only band).

## Thermal / thermomechanical fatigue (Q3) — `thermal-fatigue-tmf.json`

- **Source:** *Compilation of LCF and TMF Data of SX or DS Superalloys Used for Blades and Vanes.*
- **URL:** https://zenodo.org/records/7476962 · **License:** CC-BY-4.0 · **Retrieved:** 2026-07-01 (SHA-256 prefix `d12e6330`).
- **Slice taken:** all 450 rows with a numeric cycles-to-failure (material, SX/DS, test type, Tmax/Tmin, mechanical strain range %, Nf) across 21 Ni-base superalloys (CMSX-4, CM247LC, GTD-111, René N5, SRR99, AM1, …), parsed from the source XLSX.

## Creep (Q1) — `creep-alloy617.json`

- **Source:** *Draft ASME BPV Code Cases … for Use of Alloy 617*, INL/EXT-15-36305 Rev.2, R. Wright (Idaho National Laboratory / US DOE).
- **URL:** https://inldigitallibrary.inl.gov/sites/sti/sti/Sort_54910.pdf
- **License:** US Government work (DOE/INL) — **public domain**. **Retrieved:** 2026-07-01 (SHA-256 prefix `7bb13dd4`).
- **Slice taken:** 306 experimental creep-rupture records (temperature, stress, rupture life, creep strain) from Huntington, ORNL, INL, KAERI, ANL, CEA, GE, Cook, SMC. **NIMS-sourced rows are EXCLUDED** (NIMS prohibits redistribution). Larson-Miller master curve fitted (C=20, R²=0.975) for remaining-life estimates.

## Vibration on rotating equipment (Q6) — `vibration-bearing.json`

- **Source:** *Vibration and temperature dataset of a bearing run to failure*, Mendeley Data, doi:10.17632/5hcdd3tdvb (v6).
- **URL:** https://data.mendeley.com/datasets/5hcdd3tdvb/6 · **License:** CC-BY-4.0 · **Retrieved:** 2026-07-01.
- **Slice taken:** per-snapshot **time-domain features** (RMS, kurtosis, crest factor, peak, bearing/ambient temperature) computed from a fixed 150k-sample window of 12 hourly files spanning the 2022-06-20→06-26 run-to-failure. **Raw waveforms are not committed.** Condition is assessed by RMS/kurtosis trending + temperature rise (standard rolling-element diagnostics); displacement-fatigue is not applied to acceleration signals.