[family-assembly] Add three_jaw_scroll_chuck

[jiashunwu1030-oss]

Family

• Family: three_jaw_scroll_chuck — family-assembly
• Closes [family-assembly] three_jaw_scroll_chuck #81
• A nine-body parametric RÖHM DURO-M geared scroll chuck: one-piece body, rear cover, scroll plate (Archimedean spiral thread on the face, straight-bevel crown ring on the back), three outward-stepped BB jaws engaged in the spiral through underside arc teeth, and three straight-bevel radial drive pinions meshing the crown ring.

Official manufacturer references — copied from Family Issue #81

Only the two primary review references are reproduced here; the complete reference set remains in Issue #81.

RÖHM DURO-M product and cutaway reference (the cutaway on the right page shows the drive train this PR models: key → radial bevel pinion → bevel ring on the scroll back → spiral → jaw teeth)

Primary DIN 6350 dimensional drawing and table

---

Generated — benchmark views (what the model sees)

Image below — preview_views.png: four benchmark viewing directions used by the model.

---

Generated — easy / medium / hard samples

Image below — preview.png: easy samples are the top row, medium samples are the middle row, and hard samples are the bottom row.

---

Extremes — smallest & largest sampled instances

Image below — preview_extremes.png: the smallest sampled instance is the top row, and the largest sampled instance is the bottom row.

---

Three-view + iso — hard example

Image below — preview_hard_zoom.png: front/side/top/iso of a hard example (the benchmark views are all diagonal isos).

---

Per-part renders (assembly evidence)

Image below — preview_parts.png: every component in isolation next to the assembly — body, cover, scroll plate from both sides (spiral face / crown ring), jaw from both sides (steps / underside arc teeth), bevel pinion.

How it's parameterized

• catalog_index is discrete and maps one-to-one to all 12 official DIN 6350 cylindrical-centre-mount catalogue rows, sizes 74–630.
• Parameters carry the drawing symbol (repo convention [docs] dimension-table columns must be named physical-quantity + drawing symbol (height_G, not G) #100): chuck outer_dia_A, register_dia_B, register_depth_C, height_D, bore_E, bolt_circle_F, mount_thread_G, key_square_K; jaw BB jaw_length_A … jaw_step_H, including the previously unused BB rows jaw_serration_D (drives the scroll thread pitch) and jaw_tongue_E (drives the T-flange height).
• refine() copies the selected row's values verbatim; no dimensions are mixed across rows.
• jaw_open_fraction is the only continuous operating-state parameter: clamp_d = grip_min_A1 + jaw_open_fraction × (grip_max_A1 − grip_min_A1).
• Easy covers indices 0–3, medium 4–7 and hard 8–11; coverage enforces all 12 rows.
• The internal drive train derives per row in part.py:_layout (mirrored in check()): crown tooth count Z_w solves the axial budget and rounds up to a multiple of 3 (so all three pinion meridians are in phase), module m = 2·R_ring/Z_w, tan δ_wheel = Z_w/Z_p with Z_p = 12, shared pitch-cone apex on the chuck axis. Z_w ranges 45–84, δ_wheel ≈ 75–82° across the rows. Details and every proportion: NOTES.md.
• check() also bounds nose collision (clamp_d ≥ 0.16·jaw_width_B), guideway engagement (≥ 30 % of jaw length) and spiral engagement (≥ 2 pitches); all 12 rows stay feasible over their full published A1 range.

Engagement verification (all 12 rows × A1 extremes)

Measured with OCC boolean/extrema probes on every row at jaw_open_fraction 0.15 and 0.85:

• 9 solids everywhere, each compound member exactly one solid;
• pairwise intersection volume = 0 for all 36 body pairs (no interpenetration);
• spiral ↔ jaw-teeth interleave gap = the design running clearance (0.20–0.54 mm by size);
• crown ↔ pinion mesh backlash 0.18–1.6 mm (≈ 0.2–0.4 module);
• jaws overhang the body OD only at large A1 openings, which the catalogue's own chucking ranges imply.

bench2 validate three_jaw_scroll_chuck → PASS (12-row coverage, distinct difficulties, 9 non-degenerate solids per instance).

Visible modeling simplifications and constraints

• Exact catalogue data: 12 coupled product rows, chuck A/B/C/D/E/F/G/K dimensions, mounting-hole count, BB jaw A/B/C/D/E/F/G/H envelopes and published A1 gripping ranges.
• Bevel teeth are straight with planar flanks and computed backlash (no involute/octoid curvature); the scroll thread is rectangular in section; jaw teeth are concentric arc segments at the local spiral gap radii (the arc-vs-spiral deviation is inside the side clearance, per size).
• The rear DIN 6350 register is a recess cut into the back (per the catalogue section view); mounting holes are blind from the back, so the front face stays clean.
• G threads are nominal cylindrical holes; no false helical detail.
• Cover screws, pinion retaining details, lubrication features and edge breaks below catalogue resolution are not modelled. NOTES.md lists every derived proportion.

Checklist

• bench2 validate three_jaw_scroll_chuck passes locally
• preview.png, preview_views.png, preview_extremes.png, preview_hard_zoom.png, preview_parts.png regenerated from the current geometry and manually inspected
• family.json declares "solids": 9; every sampled body is non-degenerate
• all 12 official catalogue rows are reachable and remain coupled
• every PARAM_SPEC.source and check() rule cites the catalogue or an honest proportion/operating-state derivation
• the family directory contains exactly the package deliverables (source + NOTES.md + 5 renders)
• final commits are DCO-signed and the PR diff touches only designs/three_jaw_scroll_chuck/

Sources used

• RÖHM DURO-M product page and mechanism description: https://www.roehm.biz/en/duro-m/
• RÖHM Lathe Chucks catalogue: https://www.roehm.biz/fileadmin/user_upload/produkte/PG3_Handspannfutter/en_03_Lathe_chucks_FINISH_231121_en.pdf
  • DURO-M cutaway and operating principle (spiral ring + radial bevel drive): PDF p.19 / printed pp.3034–3035
  • DIN 6350 product rows and A1 chucking ranges: PDF pp.20–22 / printed pp.3042–3044
  • Cylindrical-centre-mount dimensions: PDF p.26 / printed p.3048
  • Outward-stepped BB jaw dimensions (incl. serration D / tongue E bands): PDF p.32 / printed p.3060
• Scroll-plate spare part showing the spiral face + bevel-ring back: https://www.carbideanddiamondtooling.com/Scroll-Plate-for-3-Inch-Scroll-Chuck--ID-658-7-887-303_p_423521.html
• Scroll chuck mechanism overview: https://en.wikipedia.org/wiki/Chuck_(engineering)

[github-actions]

Welcome, and thanks for the family PR! Three things that answer most questions:

• No checks running? A maintainer must click Approve workflows once for your first PR — pending, not ignored.
• Red gate? The message tells you the fix: click Details, or read the bench2 validate report comment posted after each run. Reproduce locally with uv sync && uv run bench2 validate <family> (debug help).
• Scope: designs/<family>/ holds only the package files (reference drawings/photos go in the family issue), and the PR checklist must be fully ticked. Not done yet? Mark the PR as a Draft.

中文:红 ✗ 点 Details 或看 validate 报告评论;本地 uv run bench2 validate <family> 复现;参考图放 issue 不放 PR;没做完用 Draft。入门 → GETTING_STARTED.zh

[github-actions]

bench2 validate

three_jaw_scroll_chuck

  ✓ family.json: keys + base_plane valid
  ✓ pieces: build(24 params) + PARAM_SPEC + check present
  ✓ PARAM_SPEC: 24 params, all entries complete
  ✓ easy: 4/4 seeds sample+check+build+execute clean
  ✓ medium: 4/4 seeds sample+check+build+execute clean
  ✓ hard: 4/4 seeds sample+check+build+execute clean
  ✓ coverage: catalog_index reaches all 12 declared values (standard table fully covered)
  ✓ difficulty separation: difficulties produce distinct programs
  ✓ geometry novelty: 12/12 unique shapes (0% duplicate)
  ✓ solids: every instance non-degenerate, 9 solid(s) each (declares solids=9)
PASS — designs/three_jaw_scroll_chuck

Previews: the previews artifact on this run.

[HaozheZhang6]

Superseded — all findings (including the jaw-to-jaw collision bound, the OD overhang, and the missing NOTES.md) are consolidated into my changes-requested review of 07-30 below. Act on that one.

[HaozheZhang6]

(Consolidates my 07-29 comment review — one review to act on.)

• Dimension check — does every spec.py number trace to the cited datasheet? Yes: all 12 rows match the DIN 6350 table cell-for-cell (catalogue p.3048 — the same drawing attached in this PR), including jaw length A = its H row. What's wrong is the typed table in issue #81: it contradicts the drawing attached right above it — "height" (50…260) appears in no row of that table (D is 32.5…135), "bolt circle" is off on all 12 rows (58 vs 63 … 580 vs 586), "center bore" on 7 of 12 (12 vs 15 at size 74 … 210 vs 240 at 630). Fix the issue body; keep the spec, and name its columns physical-quantity + symbol per #100 (height_D, bore_E). Also ship NOTES.md — REVIEWING.md §4 requires it on a table-driven family.

• Assembly validity check — the 9 bodies are separate parts: they may touch, but no two may share volume, so I measured every pairwise intersection volume (must be 0). Non-zero at every size:

  • body∩jaw 4,876–73,401 mm³ (~12–15 % of a jaw): the guideway .box() at part.py:252-256 keeps the default centered=True, which centres z too — the slots span z ∈ [−1.5·guide_depth, −0.5·guide_depth], buried inside the body, never opening the front face (also why your front views show no radial slots). You already use centered=(True,True,False) correctly at part.py:74 — single slip. Also bound travel vs slot length in check(): at size 400 clamp 191 the jaw overruns the slot's outer end.
  • body∩scroll 1,122–105,941 mm³: scroll ID max(1.12E, 0.19A) < cavity ID max(E+0.08A, 0.22A) on all 12 rows (E/A ≤ 0.38), so the plate always overlaps the central sleeve.
  • scroll∩pinion ≤753 mm³ (the 12-gon's +Z vertex pierces the plate underside); at 630, body∩cover = 78,612 mm³ — register B=545 exceeds the cover disc 0.84·A=529 (B/A runs 0.70–0.865, so the proportion doesn't dominate the catalog column; assert it in check()).
  • jaw∩jaw at small opening: check() bounds jaw_open_fraction to [0,1] but never relates clamp_d to jaw_width — jaws collide when clamp_d ≤ 0.577·jaw_width (reachable in hard, row 8 f ≲ 0.105: 6,170 mm³ measured). At the other end the jaws overhang the chuck OD (your own bbox column: 425.5 on a 400 chuck). Bound both in check().

• Mechanism check — a scroll chuck works as: square key turns a pinion → its bevel teeth mesh a crown ring on the scroll's back → scroll rotates → the Archimedean spiral on its front engages rack teeth under each jaw → three jaws feed radially in their guideways together. So each interface must exist and be engaged (interleaved or touching — not floating, not fused). None is:

  • scroll→jaw: four concentric broken rings vs six straight grooves pitched by jaw_length, floating 0.6–4.9 mm apart — concentric rings give zero radial feed. Make it one coarse spiral + concentric-arc grooves at the same pitch, interleaved 1–2 mm.
  • pinion→scroll: the scroll's back is smooth, nothing to mesh — add a toothed ring at the pinion radius, tangent with clearance.
  • retention: jaws sit in open slots with nothing holding them axially (needs a two-tier slot + T-foot); the scroll floats 33–80 mm from the cover that should sandwich it; pinions have 1.3–6 mm radial air, no journal fit.

Enclosed scroll/pinions are invisible in all four benchmark views — whether internals should be scored bodies is a maintainer call.

[HaozheZhang6]

Recommend take 1min to read: https://en.wikipedia.org/wiki/Lathe

[BenchCAD]

Pushed a rework of the internals (and updated the body to match). The previous scroll plate carried four broken concentric rings on a smooth-backed disc, and the pinion was a 12-sided prism — concentric rings give zero radial feed and there was nothing anywhere to mesh, so the modelled chuck could not work. On a real scroll chuck the drive is: key → radial bevel pinion → bevel crown ring on the scroll's back → flat Archimedean spiral on its face → arc teeth under each jaw (the DURO-M cutaway on catalogue p.3035 shows exactly this; a scroll-plate spare part shows the two faces well: https://www.carbideanddiamondtooling.com/Scroll-Plate-for-3-Inch-Scroll-Chuck--ID-658-7-887-303_p_423521.html).

What changed: the scroll face is now a true Archimedean thread (pitch taken from the BB jaw table's D band), its back carries a straight-bevel crown ring, and the pinions are straight-bevel gears meshing it with the shared pitch-cone apex on the chuck axis (tan δ = Z_w/Z_p; Z_w solves the axial budget per row and rounds to a multiple of 3 so all three pinions are in phase — Z_w 45–84, δ ≈ 75–82° across the 12 sizes). Jaw undersides carry arc teeth at the local spiral gap radii, which reproduces the real ⅓-pitch stagger between jaws 1/2/3. Also fixed from the earlier review: guideways are two-tier T-slots that actually open through the face (the old cut was z-centred and buried inside the body), the scroll is sandwiched with running clearance instead of floating tens of mm, pinions seat in stepped journal bores, the DIN register is a recess rather than a boss, mounting holes are blind from the back so the front face stays clean, and the jaw noses are narrowed per the BB plan view — which is also what lets three jaws meet at clamp_d = 2 mm without touching.

Measured on all 12 rows at both A1 extremes: 9 solids each, zero pairwise intersection volume across all 36 body pairs, spiral↔jaw interleave gap equal to the design running clearance (0.20–0.54 mm by size), crown↔pinion backlash 0.18–1.6 mm. bench2 validate passes. Spec params renamed to quantity+symbol per #100 (outer_dia_A, bore_E, …) with the BB jaw D/E rows added, and NOTES.md now documents the symbol map and every derived proportion. The old instance-map/exploded/jaw-motion attachments showed the ring-track internals, so I removed those sections; they can be regenerated from the new geometry if wanted.