Skip to content

P92 external source admission audit

Decision

Do not convert a grade name into a shared calibration identity. Keep the published 600/650 C dwell response, pure-fatigue reproduction and life-table holdout as distinct research evidence. Engineering interaction life still needs compatible rupture, fatigue and dwell validation data and an explicit failure definition. A numerical failure or solver cutback is not a physical rupture test.

Additional open reference

Imperial thesis, DOI 10.25560/60640, Prediction of long-term static and cyclic creep rupture and crack growth of grade 92 steels under different stress states. Public PDF SHA256: 57abdf756ac1198b3aab53d6b9ebe2e5043976401b49cb8e54d90e53d783c273.

Sections 3.2 and 4.4 identify the fatigue cohort as ex-service parent material (540 C, 22,000 h); cross-weld creep specimens belong to another cohort. Two fatigue tests were scrapped after thermocouple detachment. Neither the cohorts nor those invalid tests may be silently pooled with the existing references.

Visual checks of Table 5-4 (PDF p114 / printed p89) and Table 8-1 (PDF p175 / printed p150) reveal inconsistent long-term Norton coefficients: the values 3.8e-18 and 1.7e-23 are associated with opposite temperatures, while the exponents retain their ordering. Do not choose a table by which fit looks best; the constants require clarification before constitutive use.

Table 4-3 (PDF p94 / printed p69) was also visually checked. It provides an independent service-aged fatigue cohort for future transfer assessment, not same-batch validation. No thesis figures or full text are redistributed here.

Chapter 5 also describes a parent-metal creep dataset pooled from more than 150 tests with different heats/batches. Its aggregate fit cannot be relabelled as the service-aged Mat02 fatigue cohort. It remains useful for a declared cross-cohort transfer study, with material condition and scatter retained.

Industry guidance and bounded transfer

EPRI 1019778 (2010), PDF pages 55–56 (printed 5-1–5-2), was visually reviewed. SHA256: 7325f9a957d4fa3a8b56393e056761a2e13c58adce731f67ef9dc886fe35d97d. It reviews data requirements and pre-assessment, not a P92 material card. Our policy: same heat is preferred where available, not mandatory for every research comparison. Explicit cross-source transfer can proceed; silent pooling, erased pedigree, or calling transfer a same-condition validation cannot. This avoids making a data-quality preference an indefinite software blocker.

What proceeds without those constants

  • Generic prescribed-temperature input with committed-state/time semantics.
  • Co-located accepted stress, strain, temperature and state histories.
  • Public-data reproduction with explicit source/cohort and failure definitions.
  • Domain rejection and restart tests independent of fitted material values.

What remains a data gate

One traceable target material condition, compatible pure-fatigue and creep rupture observations, strain-rate and environment ranges, and independent dwell-life observations. Calibration and validation must remain separate. The 2019/2020 duplicated observations already identified in the life-screen audit do not supply an independent validation cohort. A local source-access failure or a paid standard does not justify inventing missing numeric data.

Calibration entry audit

The research life screen now validates in-memory inputs as well as file reads. Duplicate/overlapping split IDs, a missing holdout, and invalid observed lives cannot bypass the data gate through screen(data) or calibrate(data). The existing held-out predictions remain unchanged. The life, pure-fatigue, cyclic-reference and CI-routing regression passed 87 tests; this is software verification, not 87 new experimental observations.

Two-rate follow-up (2026-11-07)

Seifert, Schlayer and Maier, FVWHT 2025, pp. 36-47 is openly licensed CC BY 4.0. The retrieved PDF SHA256 is 3db99ebff289aa4ec91c67320b6d957d23f0776db23672375856e4f20ac256af. Equations on printed pp. 37-38 and the scope on p. 39 were visually checked.

The two-rate Chaboche-Jiang formulation distinguishes long and short response times, with kinematic recovery and ratcheting control. It offers a mechanism candidate for paths that a single creep law misses. However, its reported 600 C cohort is BED/519n, not the BAM 620 C cohort; complete numerical parameters are not supplied in this contribution. The law describes primary and secondary creep, not tertiary rupture. Do not invent coefficients or call its response curves an independent lifetime prediction.

The same contribution compares full viscoplastic FEM with elastic-FEM-driven local postprocessing, including a Neuber approximation. That is an alternative approximate assessment route, not an equivalent faster solver. Before adopting it, use co-located histories and compare stress redistribution and accumulated strain against nonlinear FEM. Missing parameters, source normalization and applicability remain explicit gates; no new production material follows from this literature review.