Learnings
Not a green on 2, 4, or 6. v3 column stays empty until a third article exists.
Numbers from locked cards: PIPELINE.md, FOIL_RIG.md, ROSS_CLAMP.md, aero/cfd/stage/RUN.md, cad/stator/README.md, cad/rig/BOM.md.
Findings (v1 / v2 / v3)
| # | Topic | v1 (foil-rig freeze, square plenum, integral catcher) | v2 (gallery stator + Ø12 through-path + clamp) | v3 | Verdict |
|---|---|---|---|---|---|
| 1 | Shaft assembly | Stepped stick: midspan Ø14, integral Ø16 shoulder, one-piece Ø22 catcher. 6001 ID 12 cannot pass either land. Foil ID 12.03 cannot pass Ø14/Ø22. Cannot assemble. | Ø12 through-path J1→Z=97. Two-piece T1_CATCHER_CLAMP Ø22×4 after the shaft is in. 6001 then snap/collar then 7001 CE/HCP4A from the cold end. Foil IDs stay 12.03. |
— | v1 bad. v2 good. This was a hard kill on the article. |
| 2 | Inlet feed | Single pipe into a box plenum. Dump scroll r 46→66 is exhaust, not feed. One inlet loads near vanes and starves the rest. Pitch CFD assumed uniform 25/650. | 360° gallery R=2.10, rounded channels, one +X boss Ø8. 4× Ø5 at 90° powder-escape (Forge signed empty-after-print). Housing OD 95, 0.228 kg. Not in the 2D pitch. | — | v1 bad. v2 CAD good, unproven in 3D CFD. Highest remaining aero risk on the real ring. |
| 3 | Nozzle Euler (CoolProp-on SU2) | LUT/DATADRIVEN NaN at iter 2–22. Init ρ 135.147 vs table 135.919. | Giles CENTRIPETAL, h0 1158 held. C 201 vs 208, α 70.7 vs 75, ρ on 22.19. Residual short of −8. Metal 19 / TE 0.35 held. | — | LUT bad. CoolProp+Giles good enough to move. 3% C miss is residual, not a new vane family. |
| 4 | Mixing-plane stage | Not run. | Frozen outer 94: Δh0 +56.5 vs Euler 40.4 (swirl-high, C_out 266), dump P 14.02 vs 13.89, nozzle C 181 vs 208, α 69.4 vs 75, rotor C 266 vs 120. Older 200-iter card was C 201 / +51. Continue flipped Work −7 (rotor adding h0) — killed. rms −2.17/−2.03 vs −8. 61 is the old mis-labeled static, not the freeze comparator. | — | +56.5 vs Euler-40 is swirl-high. Rotor C_out is bad. Do not iterate that restart. Viscous still NaN. Not a green on 6. |
| 5 | Wheel / mean-line | 48.00 / 13 / finish TE 0.40, PR 1.8, Δh0 75.9, η_tt 0.802. Disk cent ur 0.0148 vs Stage 0.019. Soak 0.20. | Unchanged. Axial and jet-stack families already rejected. | — | Good. Do not reopen the machine class. |
| 6 | Catcher / ROSS | One-piece 282 Ø22×4 on 0.148 kg stick. Paper first pair 25.5k. | Print 0.123 + 17-4 clamp 0.008 = 0.131 kg; + T1_7001_IR 0.0063 kg (named B3DO inner ring, not balls+outer) at Z=92.0. First pair RT 25.2k at k=1e6 / 54.7k at 5e6. 24.1k / 52.4k (+0.019 complete-bearing cap) is history. No-7001 pair on this mesh is history. 20k under on paper. UNSIGNED. 17-4 H1025 is done above ~300 C — 20k cold only. 650 C column is 282-beam paper, not a clamp cert. Do not invent inner+balls rpm. |
— | 20k cold paper is consistent. Hot clamp is bad. Residual cert still required. Do not green 2. |
| 7 | Case | 17-4 bar, CASE_OD 70, hot flange 96, ~19 mm wall over Ø32 foil bore, 2.948 kg. Lab block, not hoop-min (25 MPa membrane on r=36 wants a few mm of 282). | Same alloy. Pocket Ø28 for clamp lugs, 2.914 kg. Not thinned. One-piece. | — | Fine for a lab housing. Bad if you wanted a light can. Do not thin for the 20k article. |
| 8 | Quote / AM | Velo pack Z=72 / print TE 0.55, stub volute, no catcher. HIP 1175/150/4 h + 1135 + 1010+788. No AFM on TEs. | Quote hashes frozen. Ø12 land is post-HIP machine, not a Velo resend. Clamp is 17-4 bar, not 282 print. Gallery 4×Ø5 HIP-emptyable. | — | Quote freeze is good. Do not mix clamp into the 282 heat. |
| 9 | Coupon / Sy | — | Blank Ø12×85 Z ×6 + 2 flats. Not a 650 C AM Sy. 550/650 mass gain missing. | — | Blank is good. Cert is missing. Do not green 4. |
| 10 | CFD process | Ideal-gas / LUT as the sign. | In-process CoolProp HEOS. Giles n/t/s, not Cartesian. Mixing-plane inner 1 was cold; 20 inner took. Blind continue flipped the stage. | — | CoolProp-on is good. Blind restarts are bad. Freeze fields that hold Δh0. |
| 11 | Foil journals | Cans Ø32, film 0.015, ID would not pass v1 steps. | Foil IDs 12.03 locked. Shaft Ø12 through the cans. Eddy at Z=66 still sees Ø12. k 1e6–5e6, 20k under first pair 25.2k (T1_7001_IR 0.0063). |
— | v2 good. Do not grow IDs. Do not treat foil k as 5e7. |
| 12 | Axial locate / thrust | SKF 51101 20–40 N. Catalog 9k / 13k — 20k is not a spin plan on that part. | 7001 CE/HCP4A 12×28×8, grease attainable 98k, preload 17–40 N. Pocket Ø28. Still not 25 MPa. | — | 51101 at 20k is bad. 7001 is the catalog-legal locate. Do not green a hot machine on either. |
| 13 | Trip | Written ID ≥400 C. Not wired. 0.20 not trip-safe. | Same. Type N Ø1.5 (ID A/B, RIM, GAS) on paper only. Cold 20k does not wait. | — | Law is good. Hardware is missing. Do not grow the 0.20 to “fix” trip. |
| 14 | AMB / 80k | Parked. Bare Ø12 cannot make ≳5e7 (pole area). | Still parked until a 20k dwell exists. Sleeve target: pole ~22, gap 0.15, ≳5e7. | — | Park is good. Electrify after the spin, not instead of it. |
EE freeze (Magneto, 2026-08-31)
- SKF 51101 is dead. Axial locate is 7001 CE/HCP4A at Z=88–96, pocket Ø28, preload 17–40 N. Not 25 MPa.
T1_7001_IR0.0063 kg on the beam. First pair paper 25.2k at k=1e6 — UNSIGNED. 20k still under. Chip 2 OPEN.- Foil IDs 12.03. J2 can Z=71.00. Ø12 through-path. Two-piece clamp.
- Chip 3 OPEN (Type N not on the case). Chip 5 PARKED.
BUY.mdcompiled — do not buy.- Rig hash (B3DO, live
public/cad/cooky_t1_foil_rig.step):c7235b0f8d7e6a274af1c768489bafcda689488a9681be4d23a536be660f18d5.
P0-1 / P0-2 / P0-3 stay accepted review findings — do not delete, do not soften. P0-3 stays the 51101 catalog finding; the 7001 pick is the live locate (already the note under P0-3). No more EE geometry on this article.
Redesign, ranked by probability of actual impact
What we would change on a complete next article (v3), not a doodle. Probability is of moving a locked number (assemble, C, Δh0, first pair, hot soak) — not of looking nicer.
| Rank | Change | Why | Impact | Apply when |
|---|---|---|---|---|
| 1 | Keep Ø12 through-path + split catcher (do not revert v1 shaft) | v1 could not be built. Everything else is fiction until the stick goes through the cans. | Certain | Already v2. Never go back. |
| 2 | 3D inlet gallery in the CFD (or a 360° inlet BC that is not a 2D pitch) | v2 CAD feeds all 19 vanes. The pitch still assumes uniform 25/650. Maldistribution is the failure mode Kevin called on day one. | High | Before signing ṁ or vane TE on the real ring. |
| 3 | Mixing-plane / rotor work (solver first, metal second) | Frozen Δh0 +56.5 vs Euler 40.4 is swirl-high (C_out 266) on a field that is not converged. Rotor C_out 266 vs 120 is not. Continue already flipped the sign of Work. 61 is the old mis-labeled static, not the freeze comparator. | High (on Δh0/C_out) | Next CFD. Recut 13 / 0.40 / Ø48 only if a clean residual still dumps C_out ≫ 120. |
| 4 | Hot catcher in 282 (or drop the runner for a hot soak) | 17-4 clamp is 20k cold. H1025 is finished above ~300 C. A 650 C article cannot use this runner. | High for a hot machine, zero for the 20k cold spin | Only if the next article soaks 650 C. |
| 5 | Viscous ECM wall (ks signed, not 5 μm label) | Euler nozzle is 201/208. SST NaN'd at iter 0. η_tt 0.802 cannot be signed inviscid. | Medium-high for green 6 | After a mixing-plane field that does not flip. |
| 6 | Case as a real 25 MPa vessel, not a Ø70 bar block | Wall is locational stock. Fine for 20k. Wrong mass/OD if this becomes a packaged stage. | Medium for a product, low for T1 | v3 package, not the foil rig. |
| 7 | Parametric vane β / throat (N=19, metal 73°, α2=75°) | Nozzle Euler already 3% on C. Mixing-plane C 181 is the dirty field, not a proven throat miss. | Low-medium | Only after rank 3 plateaus with residual < −8. |
| 8 | Micro-feed nests around the ring | Printable but Forge will not sign a nest we cannot empty. 4×Ø5 gallery already has a gas path. | Low | Do not. |
| 9 | Sleeved AMB (pole ~22, 0.15 gap, ≳5e7) | Foil 1e6–5e6 crosses before 80k. Bare Ø12 AMB sits on ~1e7. Sleeve is the 1.8× split. | High for 80k, zero for the 20k cold spin | After a residual cert + 20k dwell. Not this article. |
| 10 | Real 25 MPa thrust (not 51101) | 51101 is 20–40 N locate. Radial k does not hold 25 MPa. Catcher clamp is a runner. | High for a pressurized hot machine, zero for the cold foil spin | Second CAD, with the sleeved AMB. |
v3 is not opened. Rank 1–2 are CAD we already have. Rank 3 is Stage. Rank 4 is Forge if we ever go hot. Rank 9–10 are Magneto after 2 is a run. Do not recut vanes or the quote pack on this list.
CAD lock
Existing stills only — tables stay the authority. Assembly: foils in case, shaft from the hot end, clamp, 6001, snap/collar, 7001, cover.
J1→Z=97 — two-piece T1_CATCHER_CLAMP Ø22×4 after the shaft is in (not thru), foil IDs 12.03, 1 mm air to clamp, no Ø14 swell, no integral Ø16. Findings 1 + 11.
cad/rig. Finding 2.
Blank slate (if T1 did not exist)
This is not a v2 delta. It is what Hardwar would build instead of a 48 mm / 80k / foil / 282-stick that also wants 25 MPa / 650 C.
The core mistake was one article doing three jobs: a 20k foil spin, an 80k rotordynamic target, and a 25 MPa / 650 C aero stage. Those want different shafts, different bearings, and different materials. v1/v2 keep gluing them.
| If starting over | What, specifically | Why this beats T1 |
|---|---|---|
| Two machines, not one | (A) Cold spin rig: steel/IN718 shaft, dummy or bolted wheel, split 17-4 housing, balls or foils to 20–40k. (B) Hot aero article: 282 wheel+nozzle only, no foil cans, no 17-4 clamp, real thrust TBD. | Catcher-in-the-print, 17-4 at 650 C, and 51101-as-thrust are all symptoms of the glue. |
| Split housing from day one | Horizontal split or a two-flange case (hot gas can + cold bearing cartridge). Never a one-piece through-bore with foil IDs of 12.03 and features on both sides. | v1 was unassemblable. v2 through-path works; a split case is how every turbocharger avoids that class of bug. |
| Wheel shrunk or bolted on a machined shaft | Print the 48 mm / 13-blade wheel as a disk, HIP, ECM TEs, then shrink/bolt onto a ground Ø12 steel or 718 shaft. Catcher / 6001 / 51101 live on the shaft, not in the 282 print. | We HIP'd a stick we then turned to Ø12. Quote Z=72 already admitted the print should end at the wheel. |
| Same U, bigger / slower | Keep U2 ≈ 201 m/s. 80 mm tip @ ~40k instead of 48 mm @ 80k. Same Wright-class work. Foils, catcher, and residual all get easier. Print is still 282. | 80k foil + G2.5 at 0.044 g·mm is the hard shop problem. We have not even spun 20k. |
| Or commit to 80k bearings first | If 48 mm / 80k is non-negotiable: size foils for 80k (or sleeved AMB) on day one. Do not learn 20k catcher-only and call U=201 a later-stage label. | T1's 20k article does not validate the mean-line speed. |
| Inlet scroll first, dump second | Design the 360° feed (or a real inlet volute) before the vane ring. Dump collector is easy. Starved vanes are not. 3D CFD of that gallery before a 2D pitch. | v1 dumped a box plenum. v2 CAD has the gallery; CFD still assumes uniform pitch. |
| Keep single-stage radial PR 1.8 | Do not go axial. Do not go a jet stack. Do not go 200 kW. Sandia/Wright class is the right machine. | Already rejected. Blank slate does not reopen it. |
| Gas-path 282 only | Shaft and catcher not in the 282 heat. Clamp/hot runner in 282 only if the article is actually hot. Case stays 17-4 or 316 for cold; a 25 MPa can is a different part. | Mixing 282 stick + 17-4 clamp + 650 C paper is how we got a clamp that cannot soak. |
| Thrust as a named problem | Balance piston / dummy piston / real gas thrust. Never 51101 20–40 N with a catcher disk labeled “not thrust.” | 25 MPa × eye area is orders above 40 N. |
| Type N in the first cut | Bosses and Ø1.5 holes in the first case, wired before the first dwell. | Traveler exists. Cold 20k does not wait, but the bosses should have been in v1 as-built. |
What I would not change: 282 HIP+HT, ECM/remelt TEs (no AFM), soak 0.20, odd blade count, CoolProp HEOS, no γ, catcher-only as a *backup* not a bearing.
Hardwar pick if Kevin wants a v3 blank slate: split the program. Cold split-case spin rig on a machined shaft (this year). Hot 282 stage as a second CAD, inlet gallery in the CFD, real thrust, no 17-4 in the gas. Keep 48/80k only if Magneto is on 80k bearings now; otherwise grow the wheel and drop rpm.
Stage (aero / ROSS)
U2 = 201 m/s at 80k, r = 24 mm. Spouting C0 = √(2 Δh_is) ≈ √(2×75.9e3) ≈ 390 m/s. U/C0 ≈ 0.52, not U/C2 = 201/208. Dixon radial-inflow peaks near U/C0 ≈ 0.70. We are a low-work PR 1.8 stage running slow on that chart, or a 48 mm wheel asked to do 80k foil work it cannot dump in 2D.
What is good vs bad (Stage lane)
| Good | Bad | |
|---|---|---|
| CoolProp HEOS + Giles n/t/s + AREA | h0 held 1158. MIXEDOUT flashes solid. Cartesian Giles never made 75°. | Residual never −8. |
| Nozzle Euler 19 / TE 0.35 | C 201 vs 208 (3%). Architecture held. | α 70.7 vs 75. Not a new vane. |
| Mixing-plane freeze outer 94 | Δh0 +56.5 vs Euler 40.4 (swirl-high, C_out 266). Dump P 14.02 vs 13.89. last_good/ on disk. 61 is the old mis-labeled static, not the freeze comparator. |
Nozzle C 181 vs 208. Rotor C 266 vs 120. rms −2.17/−2.03. |
| Blind continue | — | Work −7. Rotor adding h0. Do not iterate a +Δh0 field. |
| Viscous SST/SA | — | NaN / C~30. SU2 8.5 wall function will not take ks=5 μm. η_tt 0.802 is unlabeled Euler. |
| 2D midspan H-mesh | Cheap, CENTRIPETAL saw ~19 / ~14 blades. | No exducer (Japikse rotor exit is axial). Mesh dumps at r≈10 mm, wrap 42°. That is why C_out stays ~2×. No tip gap. No gallery. |
| ROSS clamp-on | 20k under RT 25.2k / 54.7k at foil 1e6–5e6 with T1_7001_IR 0.0063. Print 0.123 + clamp 0.008 + IR 0.0063. |
UNSIGNED. Zero-speed, no gyro, not a Campbell. 650 C is 282-beam paper. 17-4 clamp is 20k cold. |
| 0.20 tip | Mean-line label. | Not trip-safe vs thermal. Do not grow it to “fix” trip. |
Blank slate, ranked by actual aero/ROSS impact
If T1 did not exist. Probability is of moving C_out, Δh0, U/C0, or first-pair — not of a prettier pitch.
| Rank | What I would do instead | Why it beats this article | When |
|---|---|---|---|
| 1 | Two machines. Cold 20k spin on a ground shaft. Hot 25 MPa / 650 C stage as a second CAD with real thrust. | 20k foil + 80k U=201 + 25 MPa on Ø12 is why C_out, clamp soak, and 51101 are all lying. Mean-line 80k is not validated by a 20k dwell. | Program split. This article stays the cold one. |
| 2 | 3D radial-to-axial rotor + 360° gallery CFD before any 2D pitch. Exducer is axial (Japikse). Inlet gallery is the ring, not a Riemann 25/650. | Frozen C_out 266 vs 120 is a 2D dump at r=10, not a 13-blade proof. Gallery maldistribution is still unsigned. | Next CFD on this article (Hardwar rank 2–3). No recut until that residual is clean. |
| 3 | Size U/C0 toward 0.7 (Dixon) or admit PR 1.8 is a low-work wheel. Keep U2=201: ~80 mm tip @ ~48k, not 48 mm @ 80k. Same Wright PR 1.8. | 0.52 is off the usual radial peak. Growing D at fixed U drops rpm into the foil band. 48/80k is the hard shop problem we have not spun. | v3 only. Do not grow this 48 mm print. |
| 4 | Mixing-plane freeze rule. Copy last_good when Δh0 > 40 and C_nozzle > 180. Never continue that field. Relative-frame dump, not more OUTER. |
Continue already reversed Work. More iters is not a design change. | Now. Already frozen +56.5. |
| 5 | Campbell with gyro before anyone says 80k. Undamped k-sweep is a 20k paper gate. FW/BW split at 80k is a different number. | 25.2k first pair does not protect 80k. | After a 20k dwell, with Magneto k. |
| 6 | Viscous ECM wall after the Euler freeze, ks signed by Forge. η_tt 0.802 stays unlabeled until a residual exists. | SST NaN’d. Wall function vs roughness is a solver limit, not a TE recut. | After rank 4 holds. |
| 7 | Cold tip larger, or matched stator soak — not 0.20 as trip. | 0.20 vs thermal was already a kill for trip-safe. Growing 0.20 is not a redesign. | Hot article only. |
What I would not change: single-stage radial PR 1.8 (Sandia), odd blade count, CoolProp HEOS (no γ), 13 / 0.40 / Ø48 on this article, ECM/remelt TEs, soak 0.20 as a cold label, catcher as backup not a bearing.
Stage pick if Kevin wants a blank slate: same split as Hardwar. Aero article is 3D gallery + R-A exducer, U/C0 sized on Dixon, CoolProp mixing-plane frozen at +Δh0. This foil rig stays 20k cold on the Ø12 stick. Mixing-plane residual is still first here. Do not recut 13 / 0.40 / Ø48 on the learnings list.
360° gallery / full-annulus feed (Stage paper card, item 6 step (6))
Locked CAD (cad/stator/README.md): gallery circular R=2.10 mm, 360° annulus; one +X boss Ø8; powder escape 4 × Ø5 at 90°, empty-after-print — stay open on 20k cold, 25 MPa article plugs. Feed, not the dump. Quote hashes stay. Not a recut of cad/rig. Not a Velo send.
| Area (geometry, not flow) | Formula | mm² |
|---|---|---|
| Gallery passage | π R², R = 2.10 | ≈ 13.85 |
| 4 × Ø5 open | 4 × π (2.5)² | ≈ 78.5 |
4 × Ø5 open is ~5.7× the gallery passage. Empty-after-print, not a pressure wall — left open at 25 MPa they dump the ring. 20k cold: leave them open. 25 MPa article: plug. Do not treat open Ø5 as a sealed feed.
Why a 2D pitch + uniform Riemann 25/650 cannot close ring feed: one periodic pitch with a uniform Riemann 25 MPa / 650 C inlet assumes every vane already sees the same total. The day-one starve is the opposite: one +X boss into a 360° R=2.10 gallery — near vanes loaded, far vanes starved — and a 2D pitch cannot see that azimuth. Periodic 2π/19 plus a uniform total is the assumption, not the proof. 2D pitch mass is not ring ṁ: MEANLINE.md — ṁ is unsigned until a 3D gallery exists.
A later full-annulus run must report: ṁ split across the 19 throats (near-boss vs far); Pt loss from the +X boss to each throat; α spread at the 19 throats vs 75°. Until those three exist, chip 6 stays OPEN. Not this freeze. Not a mixing-plane continue. Full card: aero/cfd/GALLERY.md.
Fine 3D exists (NPOIN 61352) — unsigned, not conserved. Do not continue OUTER. Chip 6 still OPEN. Card: aero/cfd/gallery/RUN.md.
Forge (AM / metals)
What is good vs bad (Forge lane)
| Good | Bad | |
|---|---|---|
| Alloy in clean sCO2 | Chromia former. 750 C / 20 MPa / 1000 h: AM 282 0.60–0.61 vs wrought 0.48 mg/cm². Scale ~2 µm Cr2O3, internal ox ≤20 µm (Magnin). 740H 0.25–0.29 is thinner scale, not a reason to dump 282 creep. | 550/650 C AM mass gain is MISSING. 750 C RG is not a 650 C carburization cert. Pint: O2/H2O impurities move gain more than AM vs wrought. |
| HIP+HT recipe | Velo: HIP 1175 C / 4 h / rapid cool → vac 1135 / 30 min → 1010/2 h + 788–790/8 h. Two-step age (Joseph GB carbides). Do not furnace-cool HIP. Do not ASME 802 C single-step on a rotor. | 150 MPa is Deshpande fill (1185/150/3.5 h), not Velo-published. Silent P on a PO is a process miss. |
| TE finish | Print 0.55 / finish 0.40. No AFM on TEs — unconstrained AFM TE removal 0.29 mm. ECM or remelt only. | ks = 5 µm is UNSIGNED. As-built Sapphire Ra is not an ECM wall. Viscous CFD cannot take a label. |
| Powder escape | 4 × Ø5.00 at 90°, empty-after-print. First ports were too small to HIP-empty a 360° gallery. | Do not nest micro-feeds we cannot empty. |
| Quote pack | Z=72 / 0.55, no catcher, two 282 prints (wheel + ring). Catcher off the heat is the right split. | We still HIP a shaft (Z=0–72) then turn the span to Ø12. Paying AM + HIP distortion for a grind journal. |
| Disk / Sy | Wrought sheet Sy 631 MPa at 649 C vs rim VM 246 (margin on the sheet). | AM Sy after OUR HIP+HT is unsigned. EOS 658 MPa horiz @650 C is 1250 C Option-1, not 1175/150/4 h. Coupon Ø12×85 is a blank, not a cert. Sequence was quote-the-wheel first. |
| Clamp / 17-4 | 17-4 H1025 is fine as a 20k cold runner. | H1025 is done above ~300 C. Ø12 282 conducts from a 650 C disk — “cold end” Z=72 is not a thermal break. 17-4 cannot live on a hot article. |
| IN718 backup | Label only. | Not drop-in. Worse 650 C creep rupture, different HIP/HT. Do not quote it as primary. |
Powder-port state (4×Ø5): empty-after-print stays. On the 20k cold rig they stay OPEN — not a pressure wall. On a hot 25 MPa article they are plugged after HIP — 4×π×2.5² ≈ 78.5 mm² cannot sit in a 25 MPa boundary. Do not recut the gallery 4×Ø5. Do not nest micro-feeds.
Blank slate, ranked by actual metals impact
If T1 did not exist. Probability is of moving AM Sy, HIP distortion, hot-soak alloy, or a signed ks — not of a prettier print.
| Rank | What I would do instead | Why it beats this article | When |
|---|---|---|---|
| 1 | Coupon first. E21 Ø6.0×30 / GL 24 at 650 C after the exact traveler (HIP 1175/150/4 h + 1135/30 + 1010+788). Set C: 650 C / 20 MPa sCO2, 500–1000 h, mass gain. Then quote the wheel. | FEA is on wrought 631. AM Sy is the disk gate. 750 C RG does not close 4. We reversed the sequence. | Before any Velo PO. Closes 4. |
| 2 | Print gas-path 282 only. Wheel as a short disk (HIP-stable Z), nozzle ring with empty-after-print ports. Shrink or Hirth onto a ground 718/steel shaft. Journals are ground, not AM lands. Catcher/6001 live on the shaft. | HIP 1175/150/4 h on a slender stick warps. We then machine the whole span to Ø12. Quote Z=72 already admitted the print should end at the wheel. | v3 architecture. This article stays the Ø12 stick. |
| 3 | Two alloy articles. (A) 20k cold: 17-4 case, steel shaft, 17-4 clamp. (B) 650 C / 25 MPa: 282 wheel+nozzle only. Any runner that can see >300 C is 282 or 718, never H1025. | 17-4 clamp + 650 C paper on one stick is how we got a runner that cannot soak. Conduction along Ø12 282 will heat Z=72. | Program split. Same as Hardwar rank-1, with the °C cut. |
| 4 | Sign ks from a PECM/remelt coupon Ra, not a 5 µm label. AFM stays off 0.40 TE (0.29 mm kill). PECM vs laser remelt is the TE fork (Gradl/MSFC PECM analog, recast vs electrolytic). | Stage viscous NaN’d on an unsigned wall. η_tt 0.802 is unlabeled Euler until ks is a measurement. | After mixing-plane freeze. Closes the wall, not the vane. |
| 5 | Put HIP P on the PO. 150 MPa Ar as a stated Deshpande substitute, or get Velo’s number. Rapid cool, no furnace cool. Keep 1135/30 — do not skip. Keep two-step age. | Silent 150 MPa is how a traveler drifts to 100 MPa / 1250 C (Shaikh/EOS) and we lose the grain-size/creep card. | Same PO as rank 1. |
| 6 | Do not swap 282 for 625/740H on RG, and do not open IN718 without a separate 650 C rupture card. | Thinner scale at 750 C is not higher rim Sy. 282 is the γ' creep pick. | Never on this class. |
What I would not change: Haynes 282 on the gas path, two-step 1010+788, no AFM on TEs, print-fat 0.55 / finish 0.40, 4×Ø5 empty-after-print, quote cut before the catcher, coupon as a traveler (but run it first).
Forge pick if Kevin wants a blank slate: same split as Hardwar. Coupon + HIP-P on the PO before metal. Print a 282 disk, not a 282 stick. 17-4 never in the 650 C conduction path. This foil rig stays 20k cold. Do not recut 13 / 0.40 / Ø48. Do not green 4.
External review (31 Aug 2026) — what we take
Accepted (P0)
| # | Finding | Evidence on our disk | Action |
|---|---|---|---|
| P0-1 | Case and cover occupy the same metal. Review: ~29 065 mm³ common (~75% of cover). | T1-CASE cold flange is Z=100–108, Ø80. T1-COVER is Z=99.2–108, Ø80. The cover was modeled on top of a flange the case already owns. |
@B3DO new assembly rev: case ends at the mate (Z=100), cover is the plate + spring boss only. Recalc spring working length. Do not thin the case. |
| P0-2 | Euler work ≠ displayed Δh0 75.9. With U2=201, C2=208, α2=75°: Cθ2=201 m/s. Zero-exit-swirl work U2 Cθ2 ≈ 40.4 kJ/kg. Cap at |C3|=120 is ~49 kJ/kg, not 75.9. Static drop 1136−1075=61 was labeled Δh0. | MEANLINE.md table vs _eta.json 75.9. Review identity is correct. CoolProp cannot fix angular momentum. |
@Stage closed aero/MEANLINE.md at 40.42; CFD target 40.4. Frozen mixing-plane +56.5 vs Euler 40.4 is swirl-high (C_out 266), not a miss of 75.9 and not a miss of 61. 61 is the old mis-labeled static drop. Do not retune vanes to 75.9. |
| P0-3 | SKF 51101 catalog speed is below the 20k dwell. Review: SKF reference 9 000 / limiting 13 000 rpm. Preload 20–40 N does not raise that. | We named SKF 51101, Z=88–97, 20k ceiling. | @Magneto pick a thrust part whose catalog envelope covers the dwell, or drop the dwell below the catalog limit. 20k on 51101 is not a spin plan. |
Magneto’s pick is SKF 7001 CE/HCP4A (finding 12 / /ee). P0-3 stays the accepted review finding — 51101 catalog is below 20k.
Accepted (P1, do not green)
| Finding | Take |
|---|---|
| Shaft is hollow Ø12 / Ø10. Snap groove Ø11.50 leaves 0.75 mm wall. Solid-Ø12 I is ~1.93× this tube. | “Ø12 through-path” hid the bore. @Stage ROSS/FEA must use the tube + groove + clamp + collar + snap + rotating washers, not a solid 0.123 kg stick. Until that hash matches CAD, 25.5k stays paper. |
| Two rotor solids in one STEP (print + finish) | BOM/inertia/collision must have one active rotor. Print-vs-finish is a compare config. @B3DO. |
| Mixing-plane outer 94 is a checkpoint, not a solution | Already frozen. Residual ~−2. Do not call it last-good as if converged. |
| Nominal 0.20 gap ≠ running clearance | Soak label only. Budget orbit + cent + thermal + foil 15 µm + catcher 150 µm at their own stations. Do not grow 0.20 to “fix” trip. |
| V2 gallery R=2.10 is a size, not a proof of uniform feed | 2D pitch cannot sign the ring. Powder ports need an operational state (plugged vs flow). |
| Coupon blanks are bars, not thin-feature witnesses | Keep blank ≠ Sy. Next coupon set needs TE/wall witnesses. 4 stays open. |
Not taken as a recut
- New blade family, faster spin, thinner case, turbulence knobs to hit 0.802.
- Treating 38 kJ/kg rotor Euler as “wrong vs 75.9” — it is near the triangle work. The reference is what is wrong.
- Origin-not-found / UI polish: site hygiene, not hardware.
Next steps (this article, in order)
| Step | Owner | Question | Done when |
|---|---|---|---|
| 1 | B3DO | Can the parts coexist? | Case/cover no volume interference (split Z=100 mate). One active rotor in the rig STEP. Spring working height declared. Sequence: foils → shaft → clamp → 6001 → snap → 7001 → cover. |
| 2 | Stage | Does the duty match the triangles? | Closed in aero/MEANLINE.md. Euler Δh0 = 40.42 kJ/kg (U2=Cθ2, Cθ3=0, r3=9 mm). 75.9 was η_ts×101 totals; 61 was static h2−h3 mis-labeled Δh0. CFD export is 40.4, not 75.9. ṁ still unsigned. |
| 3 | Magneto | Can the cold rig run at its ceiling? | 51101 is out (catalog 9k/13k). Locate is SKF 7001 CE/HCP4A 12×28×8, pocket Ø28, grease 98k, preload 17–40 N. Still not 25 MPa. 2 stays open until a residual cert. |
| 4 | Stage | Does ROSS/FEA describe this shaft? | Tube Ø12/Ø10, groove, clamp 0.008, collar, snap. 7001 rotating mass is named T1_7001_IR 0.0063 kg (inner ring only, not balls+outer). First pair 25.2k / 54.7k unsigned. Catalog 0.019 / 24.1k is comparison only. Do not invent inner+balls rpm. Still no gyro. Hash vs STEP. Chip 2 stays OPEN. |
| 5 | Stage | Is CFD a solver bug or a design miss? | Card, not a green. File aero/cfd/SOLVER_VS_DESIGN.md. Freeze +56.5 vs Euler 40.4, not vs 61, not vs 75.9. Extra is 2D exit swirl (C_out 266). Keep last_good. No continue into −7. 19 vs 13 pitch scaling. Chip 6 OPEN. SST NaN at iter 0 is a solver card, not a TE recut. |
| 6 | Stage | Does the gallery feed 19 vanes? | Full-annulus or a 360° inlet BC. 2D pitch cannot close this. Paper card aero/cfd/GALLERY.md: bore ≈13.85 mm² vs 4×Ø5 ≈78.5 mm² — areas, not a ring ṁ. Chip 6 stays OPEN. |
| 7 | Hardwar / Stage | Does any transient eat the 0.20? | Clearance budget vs time. Not a trip-safe claim. |
| 8 | Forge | Thin-feature process | Set D — TE / wall witnesses on a later plate (COUPON.md Set D), not a recut of the frozen Ø12×85 bars+flats coupon STEP. Rotor TE 0.55→0.40, stator TE 0.55→0.35, ECM or remelt, no AFM. Ra→ks — 5 µm stays UNSIGNED until that number. 4 stays OPEN until set A 650 C Sy (E21). Set D does not green 4. ks does not green 6. No PO. |
| 9 | Stage | Aero DOE | Only after steps 2 and 5. No 13-count change to force 0.802. |
| 10 | Hardwar | Cold hardware | Only after 1, 3, 4. Residual cert + catalog-legal dwell. Still no PO from this list. |
Review 31 Aug (Stage takeaways)
P0 work identity is real. Closed: Δh0 = 40.42 kJ/kg (aero/MEANLINE.md). Do not tune mixing-plane to 75.9. Frozen +56.5 vs Euler 40.4 is swirl-high; C_out 266 is swirl, not a 13-blade kill and not a 61 miss. 61 is the old mis-labeled static drop, not the freeze comparator. Chip 6 stays OPEN.
P1 hollow shaft: clamp ROSS already uses Ø12/Ø10 (BORE_M=0.010, snap Ø11.50). Not a solid beam. Current first pair RT 25.2k / 54.7k is T1_7001_IR 0.0063 kg (inner ring only). Catalog 0.019 / 24.1k is comparison only. Still not a Campbell: no gyro, no residual cert. Chip 2 stays OPEN. UNSIGNED.
Snap groove 0.75 mm wall on Ø12/Ø10: fine as a 20k cold machine land. Not an 80k 282 notch card. Not a recut.
P1 “last good” is a freeze, not a converged solution. Chip 6 stays OPEN.
Do not recut 13 / 0.40 / Ø48 on this review.
v3 stays closed. Quote and gallery hashes stay unless step 1 forces a case/cover rev.