Skip to content

ASTM Batches

Peter Pak edited this page Aug 25, 2026 · 1 revision

ASTM Batches — batch → build → process-point map

Working reference for paper figures/results, compiled 2026-08-25 (triage from the latest batch backwards). Sources: astm_specimens notes, build_registry, agent_actions (the tool-call audit trail carries the exact profile_set payloads even where the printer's job/profile JSONs haven't been rsync'd yet).

Raw curves: Agentic-SLS-ASTM repo, per-specimen source_paths column (tensile xlsx per test date; flexural for Q/R/T/U in source/2026_08_23/, not yet imported into astm_specimens). Telemetry: per-build parquets in Agentic-SLS-Telemetry/source/telemetry/<NNN>.parquet. Setup conversations: Agentic-SLS-Conversations/source/sessions/ (2026-08-16 through 08-18 sessions cover the blend campaign).

PA12 White / PA12 GF blend campaign (batches S, T, U)

Material: 25% Formlabs PA12 GF / 75% PA12 White powder blend (pure White cannot couple the 420 nm laser at any dose; the GF fraction is the absorber — see build 74 fuse test, 2026-08-16).

Batch Build Printed Job / process point
S 77 2026-08-16 "D638 Type 4 and D790 - PA12 White-GF Blend (168C Surface, 28mJ Fill, 110% Powder)" — profile authored by antigravity harness. Only samples 1–7 exist: soft-canceled as the upper stack began; laser energy raised mid-print (build 76 is the earlier same-day attempt of this job).
T 78 2026-08-17 Blend v3: 56 mJ/mm fill, 56-48 outlines, 168 °C, 130% powder, 70 min heat, no heat-up recoats. Outline override explicitly staged pre-print. Samples 1–7 lower level, 8–14 upper.
U 79 2026-08-18 Blend v4: 56 mJ/mm fill, 172 °C surface, begin-layer 176 °C + 20 s dwell, 130% powder, 70 min heat. Full stacked layout (1–7 lower, 8–14 upper). Outline-energy caveat: profile scalars say 56-48 but the legacy outline array was stale at [28, 18] and no pre-print staging override was applied (unlike T) — outlines most likely printed at [28, 18].

Batch U profile (agent_actions id 411, profile "2026_08_18 PA12 White/GF Blend v4", base ea260ec3): surfaceTarget 172, surfaceTarget2 174, heatingTargetPrint/Bed 172, printCapTemperatureTarget 174, beginLayerTemperatureTarget 176 (+20 s delay), bedPreparationTemperatureTarget 174.

Builds 74 (fuse test) and 75 (46 mJ card print) are the same-material precursors; build 76 is the aborted first attempt of batch S's job. Builds 74/79 have blank job_name in builds (recorder joined mid-build) — job identity comes from agent_actions / inova_print_sessions instead.

PA11 Onyx campaign (batches O, P, Q, R)

Batch Build Printed Job / process point
O ~65 2026-07-29 First PA11 print in dataset; 4 D638 specimens (lab numbers 10–13). Believed to be "D638 Type 4 - PA11 Onyx (186C Surface, 36mJ Fill) (2026-07-29)". Tested 08-03.
P 66 + 67 2026-07-30 D638: build 66 "186C Surface, 32mJ Fill"; D790: build 67 "D790 and Benchy and Rocket Trophy". Specimen 7 doesn't exist (numbering skips 6→8). Tested 08-03.
Q 68 2026-08-06 "186C Surface, 32mJ Fill, 32-22 Outlines". Specimen 7 broke. Tested 08-07. Flexural xlsx in 2026_08_23/ pending import.
R 69 2026-08-07 "188C Surface, 34mJ Fill, 34-24 Outlines". Tested 08-08. Flexural xlsx in 2026_08_23/ pending import.

Batches O–U are all curve-only (analyzed strain/stress trace from the TestWorks xlsx; no geometry or scalar-metric columns — peak stress is derived from the curve, modulus/strain metrics are null in astm_specimens).

Quick peak-stress reference (D638, MPa)

U 18.1 ± 14.0 (max 44.4, n=14) · T 12.2 ± 10.9 (max 35.2, n=14) · S 4.9 ± 5.0 (n=7) · R 13.9 ± 10.3 (max 36.3, n=14) · Q 10.1 ± 6.7 (n=13) · P 7.6 ± 5.4 (n=11) · O 13.8 ± 2.1 (n=4). High in-batch scatter in the blend/PA11 batches is real (stacked layouts, per-level variation), not an import artifact.

Earlier batches (A–N, PA12 GF era)

Builds ≤ 67 have job/profile linkage in Agentic-SLS-Database/source/build_to_inova_session.csv (refreshed through build 67) and full scalar metrics in astm_specimens (modulus, yield, strain at break). M is mixed-geometry (Type I + one Type IV); N is xlsx-only like the later batches; J_MB is the masterbatch variant of J.

Backup status (2026-08-25)

  • ASTM repo: committed + pushed (tensile through U, flexural Q/R/T/U).
  • Conversations repo: sessions through 08-19 + conversations.jsonl pushed.
  • Database repo: pushed through build 67 linkage. Pending: printer rsync of PrintSessions/Jobs/PrintProfiles for builds 68–80 (printer was offline 08-25); until then the v4/v3 profile JSONs exist only as agent_actions payloads (local dump: data/backups/agent_actions_2026-08-25.jsonl).
  • Telemetry repo: builds 68–80 parquet export run 2026-08-25.
  • build_summaries for 68–80: regenerate with sls-summarize-builds.

Clone this wiki locally