P92 prior-damage remaining-life audit¶
Result and provenance¶
Two public author manuscripts provide a useful sequential-loading discriminator:
- Microstructural damage mechanics-based model for creep fracture of 9%Cr steel under prior fatigue loading, Table 1, PDF p5.
- Evaluation of the effect of various prior creep-fatigue interaction damages on subsequent tensile and creep properties of 9%Cr steel, Table 1, PDF p22.
Both tables were rendered and visually checked. evidence/creep/p92-prior-damage.json
retains PDF hashes, 15 model-table observations and published predictions, plus
eight links to the experimental table. PDFs and figures are not redistributed.
The two manuscripts reuse experiments; they are not independent datasets.
Retain the differing 1143/1144 h and 186/185 h entries explicitly. Do not average
them or treat a difference as a new replicate.
The experiment identifies a normalized/tempered 105 mm OD, 24 mm wall P92 pipe, 650 C in air, prior strain amplitude 0.4%, strain rate 0.001/s and subsequent creep at 130 MPa. Its prior-life fractions use the shortest of at least three reference fatigue tests per hold condition, not a mean fatigue curve. The model table additionally includes prior LCF at 0.25% amplitude. No common heat with the existing thermally aged rupture data or 600/650 C dwell-life table is established.
Reproducible arithmetic, not an executed constitutive law¶
Run from the source root:
The author used as-received creep at 125/130/140 MPa and prior 0.4% LCF to identify parameters (sections 5.1-5.4); the table's corresponding rows are 1, 6-9. Rows 2-5 and 10-15 are the author's transfer group. This audit does not re-fit anything and does not call these an AgentFEM blind validation split.
| Published prediction group | Mean absolute relative error | Largest relative error |
|---|---|---|
| Calibration-associated table rows | 3.123% | 7.952% |
| Transfer rows, all retained | 51.956% | 326.357% |
At 50% and 70% prior C-F life fraction with 180 s holds, published predictions 550/479 h are compared with observed 129/186 h: ratios 4.264/2.575. Section 5.4 attributes the additional degradation to surface cracks beyond microstructural recovery. These are useful failure cases, not outliers to discard. Neither these ratios nor a factor-of-two plot is an engineering acceptance criterion.
Next bounded implementation¶
- Reproduce the published material-point equations with fixed parameters before any new calibration. Check equation units, initial conditions and the rupture event explicitly; an integrator failure or cutback is not physical rupture.
- Preserve calibration-associated and transfer groups. Include all high-damage failures. Measured lath width and a fraction of known fatigue life are inputs to this conditional prediction, not outputs from an arbitrary service history.
- Investigate whether a crack-sensitive damage description is needed for the failed regime. Do not invent a correction from those two points and then reuse them as independent validation. First locate additional compatible crack/history observations or restrict the model's admitted domain.
- Only after numerical reproduction, connect the local law through the existing material/state/Step interface and verify homogeneous structural response and restart. Do not create a P92-specific global solver or claim a component life allowable from this single-stress dataset.
The earlier 2017 ResearchGate landing page currently exposes only an abstract; search snippets suggesting full text were not treated as numeric evidence. The public institutional manuscripts above are the actual reviewed sources.