Feeder design review — pre-fab gate

Important: do not rigidly connect vibrating hopper and fixed trommel. Clearance/flexible connection needed.

What's decided

Tier 1 Andi's exact requirement is what Q17 specifies. RUBISCO2/Sand Separator/docs/v2/qdrum_annex/Q17_drum_feeder_interface.md (locked 2026-04-26) chooses Option B — sliding overlap with EPDM-bristle skirt:

• Chute exit 700 × 300 mm rectangular telescopes 25 mm inside the cone (PRT-80004, 900 mm ID round).
• 15 mm radial gap absorbs ±10 mm field misalignment.
• EPDM bristle skirt clamped to the chute mouth fans radially, seats against cone inner wall.
• Explicit rationale in Q17: "vibration decouples — gap absorbs all relative motion."

Why Andi missed it

Q17 lives in the Q-Drum annex; the bible §11 (feeder) doesn't reference it; the RevL fab pack doesn't either. A reader of the bible + fab pack does not encounter Q17. Documentation gap, not design gap.

Hole pattern, open area, hole size, edge distances, wear allowance, and support structure are missing. Function and strength unclear.

What is in the spec

Tier 1 spec.py:47 sets floor_thickness_mm: 5.0 for the bin floor; material A36. part_specs.py BIN-PRT-30060 enumerates only fastener holes (12× M12 through-bolts to spring/sub-frame, 14× M8 lateral, 4× M8 rear, 4× M8 funnel). No sand-pass-through perforation pattern is defined.

This raises a deeper design question I should surface for you: in Plan-D, where exactly is the screen-pass-through? The drum has the modular 4 mm 316L screen panels (bible §4). The feeder's job is to vibrate the material and deliver it to the drum; whether the feeder also acts as a coarse pre-screen is a design decision that I cannot find a clear answer to in the current spec.

• If the feeder bin floor is solid: Javier's "lámina below the feeder" requirement (meeting_notes_ordered.md) implies sand somehow reaches the deflector PRT-30070 — but if the floor is solid, sand can only spill over the chute lip, not fall through. Then PRT-30070 catches drum spillage, not feeder pre-screen.
• If the feeder bin floor is perforated: hole size must match the drum-screen size (4 mm) so what the feeder passes the drum doesn't try to pass again, and edge distances + open-area must be specified.

Andi can't tell which is intended — and neither can I from the current spec.

Tier-3 default if no decision: solid floor, treat the deflector as a spillage-and-overshoot capture device, and rely on the drum screen as the only sand pass-through stage. That matches the "feeder = transport, drum = separation" division of labour and is the simplest fab.

Maybe critical, when perforated. 1t dynamic local impact loads may deform it. (maybe more PTRs behind it help)

What's known

Tier 1 The bin floor is 5 mm A36 plate. spec.py:53 defines rear_loader_impact_plate_height_mm: 300.0 — there's a separate impact wall at the rear loading end, but the floor itself is single-thickness. The 1-ton loader-slug load case is analysed only at the spring level (slug_transient_load_per_spring_kg = 300 kg, deflection 58.5 mm of 104 mm linear range, design SF 1.78). Local plate-bending under a one-sided concentrated 1-t dump on a 5 mm × 1.65 m × ~2 m unsupported plate is not analysed.

Why Andi's right

5 mm A36 plate spans the bin width unsupported between sub-frame crossmembers. Yield is ~250 MPa; under a 1-t load distributed over a tractor-bucket footprint (~0.3 × 0.5 m), local plate stress can hit yield. Andi's "more PTRs behind it" suggestion is the right direction: more ribs / crossmembers under the floor distribute load.

Must be designed for 1-ton dump load plus drop height, not only static load. At minimum evaluate "one-sided excavator dumping" case.

What's done

Tier 1 Bible §11.2 Q4 names the 1-ton loader-slug case. spec.py:264-277 sizes the springs against this case (slug 300 kg/spring, 58.5 mm deflection, SF 1.78 vs max linear range). The spring stack is the only subsystem analysed against impact.

What's missing

• Local plate stress on the bin floor (item 3 above).
• Drop-height contribution. Bucket release height is typically 0.5-1.0 m; impulse = (load × √(2gh)) raises peak force ~2-3× over the static-equivalent slug. The current "1-ton slug" treatment is a static-equivalent that may understate peak.
• Asymmetric / one-sided dumping — the spring analysis assumes symmetric 4-point reaction. One-sided dump puts ~80 % of the impulse into 2 of the 4 springs, which roughly doubles their per-spring transient. SF for two springs at slug = 0.89 (below 1.0).

Four springs may be insufficient. Risk: excessive compression, bottoming out, poor vibration stroke.

What's specced

Tier 1 4× helical compression, OD 100 mm, wire 12 mm, free 200 mm, 6 active turns, k = 50.27 N/mm per spring, system k = 201.1 N/mm, natural f = 3.81 Hz, drive 25 Hz, isolation ratio 97.6 %. 51CrV4 chrome-vanadium, shot-peened. Static SF 3.55, slug SF 1.78. Source: spec.py:214-294. Spec source-tagged: "derived in cadquery_part.py geometry review; phase-B-stiffened to handle bible §11.2 Q4 1-ton slug. Not yet validated against a vendor datasheet."

What Andi is reading

Without the spec text he sees four cans on a CAD render and reasons from "is 4 enough?" The numbers are defensible — but the SF goes to 0.89 if item 4's one-sided-dumping case applies (see above), so Andi's "borderline" instinct is correctly calibrated to a real load case.

Add upper/lower bump stops and lateral guides for high impact.

Status

Tier 1 Bump stops, snubbers, and lateral guides are not specified anywhere in the current feeder spec. Standard practice on vibrating-screen feeders is rubber/polyurethane bump stops at the hard limit of spring travel (~95 % deflection) so a slug overload bottoms out on a controlled snubber rather than coil-on-coil. Lateral guides prevent the bin from walking sideways under asymmetric load.

Why this matters

If item 4 (one-sided dump SF 0.89) bottoms a spring out, coil-on-coil contact at peak transient creates a local ~459 MPa shear stress (already at the spring's allowable per Wahl correction at slug). Repeated coil-on-coil contact fails the spring. A stops kit limits travel to the linear range and converts overload into a controlled rubber-deformation event.

Type, RPM, excitation force, synchronization, and vibration direction missing. Function cannot be properly evaluated without this.

What's in the spec

Tier 1 spec.py:297 "OLI MVE 800/3 or equivalent", 2× counter-rotating, body 700 × 385 × 320 mm, mount plate 820 × 460 × 12 mm, 880 mm centre-to-centre spacing, 4× M16 per motor up through sub-frame motor cradle crossmembers at x = 500/1120/1380/2000.

The bible §11.4 mentions OLI 3.0-3.5 kN centrifugal force, 220V 3-phase, IP65 minimum, VFD-driven 30-80 Hz sweep. Drive frequency 25 Hz / 1500 rpm captured in spec.py:260.

What's missing in the fab pack

• Centrifugal-force rating per motor (Andi can't check kN.)
• Synchronization mode — natural frequency lock vs electrical sync vs mechanical link. With counter-rotating pairs, the standard is rigid frame sync at ~2× drive frequency above natural; this works only because the sub-frame is rigid.
• Vibration direction — counter-rotating pairs project linear vibration along the line perpendicular to the rotor axes. The current geometry implies horizontal-X linear vibration; this should be drawn explicitly with an arrow on the cover sheet.
• OEM datasheet block on the fab pack so Pedro/Marco buy the right unit.

Marco addendum (2026-05-08, WhatsApp)

And better have 1 motor instead of two. Otherwise we will have sync problems.

Q1 already decided this

Tier 1 RUBISCO2_Feeder_Q1_Motor_Placement.pdf evaluated five options A-E. Option E was explicitly the single-motor centred case. The doc named E's pros — "Cheapest. No synchronization problem — a single motor is always in phase with itself." — and rejected E on transport grounds: "Produces circular/elliptical stroke, not a linear stroke biased toward the cascade — slower material transport. Single point of failure."

Q1's recommendation was Option A (twin counter-rotating, centred transversely): "Counter-rotating twin pair produces a defined linear stroke along the long axis — drives material toward the cascade edge." The trade-off was already on the table — A traded sync complexity for linear transport.

So Marco is not introducing a new concern. He is re-weighting a known one: linear transport > sync risk in Q1 vs sync risk > linear transport in Marco's call.

What changed since Q1

Tier 1 Time constraints have killed the Apr 21 test rig — production unit goes straight to full-scale, no empirical validator before commit. That removes the path "validate single-motor on rig and switch if it works." Every choice gets validated for the first time on the production unit itself.

Failure-mode recoverability (production unit)

• Option A (2 motors) failure mode: motors drift out of sync → motion goes elliptical → transport less efficient. Recovery: tune VFD master-slave electrical sync, retighten frame, swap to matched-lot motors. Material still moves.
• Option E (1 motor) failure mode: circular stroke too slow for wet sargassum at target throughput. Recovery: not really — it's a physics constraint, not a tuning issue. Adding a second motor means redoing the motor cradle.

The frame already accommodates either

Tier 1 spec.py:313: 4 motor cradle crossmembers at x = 500/1120/1380/2000. The bolt pattern supports 2-motor (Option A at x = 810/1690) and 1-motor (Option E at x = 1250 centred). The cost of keeping both options open is zero fab change.

Forks (no test rig)

• Fork A — Option A as Q1 decided. Mitigate Marco's sync concern with: (i) FEA sub-frame torsional natural f < ½ drive f before fab, (ii) buy both motors from a single supplier batch (matched s/n), (iii) spec VFD pair as master-slave electrical-sync capable as backup, (iv) commissioning sync-verification at 25 Hz / 30 min observation. Pro: linear transport, mitigations are well-understood. Con: 2× motor cost, real sync failure mode that mitigations may not fully close on a hand-fabbed unit.
• Fork B — flip to Option E (1 motor). Honors Marco. Accept circular/elliptical transport. Pro: zero sync risk, half motor cost. Con: Q1 explicitly rejected this on transport grounds; risk is "circular too slow for wet sargassum" with no recovery path other than retrofit.
• Fork C — build frame for 2, downgrade to 1 in commissioning if sync fails. Buy 2 motors per Q1 + Fork A's mitigations. If commissioning shows uncureable sync drift, remove one motor and run as Option E with no fab change. Pro: honors Q1 decision, addresses Marco as real fallback (not hand-wave), cost of optionality = 0. Con: still buys 2 motors upfront; downgrade decision happens late, after fab.

No-Go for beach/sand operation. Larger pads, skids, or base beams required.

The design intent

Pablo's voice note 2026-05-07: feeder feet do not bear directly on sand. Drum frame and feeder frame both bolt to a common machine skid / connecting plate; that plate distributes load.

The gap

Tier 1 The connecting plate is not in the RevL fab pack. The reader of the FEEDER pack alone sees the leveling-foot pad (Ø 115 × 12 mm, spec.py:207) and reads "small pads, sandy ground = no-go" exactly as Andi did.

Forks

• Fork A — add the connecting skid as a new WI inside the FEEDER pack. Pro: closes Andi's bullet 8 in one deliverable. Con: skid is a machine-level item, scope-creeps the FEEDER pack.
• Fork B — open a Machine-Skid WI under v2-machine-combined. Pro: clean ownership. Con: another deliverable before Annex B.
• Fork C — cover-sheet note: "feeder feet bolt to a common machine skid (separate WI, in progress); feet do not bear directly on sand." Cheapest, kicks the structural design downstream.

Too slender for 1-1.5 t plus vibration. Increase size, lock them properly, protect against sinking.

What's specced

Tier 1 PRT-70015 — M20 jack screw with spherical-pivot foot (MISUMI LSCB), 4× per machine, foot pad Ø 115 × 12 mm A36, anchor hole Ø 18 mm, top hex socket AF 17 mm. Source: spec.py:203-212 + COMPONENT_INVENTORY.md.

• Bending under lateral: spherical pivot decouples lateral load → axial through the rod. Andi's "thin rod will bend" mode requires fixed-base feet, which is not what's specced.
• Axial capacity: M20 grade-8.8 minor area 314 mm² × 660 MPa yield = ~207 kN axial yield. 1.5 t / 4 feet = 3.7 kN per foot. SF ~55× axial.
• Sinking on soft ground: Ø 115 pad → 0.0104 m². 1.5 t / 4 = 3.75 kN per foot → 360 kPa contact pressure. Loose dry sand bearing capacity is 50-100 kPa. Andi is right that the current pad will sink.

The real gap

Pad bearing pressure exceeds sand allowable by ~3-7×. The connecting-plate plan (item 8) dissolves this — once feet bolt onto a 2 × 1 m skid plate, contact pressure drops by ~170× and falls well within sand allowable. Without the plate, Andi is correct.

Poor design. Sand will likely accumulate near the feet and bury them. Chute should discharge further away.

What is there

Tier 1 PRT-30070 sand deflector — 304 SS plate, 3 mm, 2300 × 1000 mm, 30° tilt, centre at world (1250, 500, −725). The Z = −725 puts it well below the bin floor; the 30° tilt redirects falling material laterally away from the leg footprint. Confirms Javier's requirement (meeting_notes_ordered.md): "metal plates below feeder AND drum to redirect sand so a tractor can collect it from the side."

What's actually wrong

1. The despiece for PRT-30070 shows the plate in isolation; no section view of the bin showing the full sand path screen → deflector → lateral exit.
2. The Y-span is 1000 mm, centred at Y = 500. If the legs sit at Y < 0 or Y > 1000, sand sliding off the deflector at 30° projects a downstream cone of trajectories that can still land near the leg footprint.

Pablo's instinct "maybe extend a bit" is the correct geometric correction.

May work, but motor loads and weld fatigue should be checked, especially at motor support crossmembers.

What's specced

Tier 1 Sub-frame: A36 RHS 50×50×3 mm (spec.py:143). 4 motor cradle crossmembers at x = 740/1140/1360/1760 with 4× M16 per motor up through them.

What's missing

Motor loads × eccentric weight × ω² generate ~3.0-3.5 kN per motor at 25 Hz drive. With 880 mm motor spacing, the inner crossmembers carry alternating ±3 kN at 25 Hz. Weld fatigue (FAT 80 typical for fillet welds in cyclic load) gives ~2 × 10⁶ cycles allowable at full stress range. 25 Hz × 8 h/day × 250 days/yr ≈ 1.8 × 10⁸ cycles/yr. Stress range must drop to FAT 36 territory or below to pass design life — that's typically ~50 MPa nominal weld stress. Not checked in current spec.

Critical. Thread locking, lock nuts, accessibility, and inspection strategy must be defined.

What's specced

Tier 1 spec.py:58: "12 × M12 weld nuts on bin underside for SUB-FRAME perimeter anchor." spec.py:165: "M12 zinc bolts down through bin floor + underside doublers into M12 weld nuts on sub-frame top rail."

Why this is a real concern

Weld nuts under continuous vibration are a known fatigue site. The HAZ around the weld nut is a stress concentrator; under 25 Hz drive × millions of cycles, micro-cracks initiate at the weld toe and propagate. Mitigations: (a) Nord-Lock washer pairs (replace pre-load loss), (b) thread-locker (Loctite 243 medium), (c) K-monel locknuts above the weld nut, (d) inspection schedule (re-torque every N hours).

None of these are currently specified. Andi's call-out is on the money.

Abrasion and corrosion are concerns. Protective coating, drainage, and replaceable wear plates recommended.

What's known

Tier 1 Most feeder structure is A36 (frame, bin walls/floor, doubler plates, spacer plates). The deflector is 304 SS. Spring is 51CrV4 with shot-peen + zinc-rich epoxy. Bin floor / frame coating: not specified.

Per memory: Caribbean sand is CaCO₃ (Mohs 3-4), so corrosion — not abrasion — is the dominant material threat. A36 in salt-laden coastal air without coating gets pitting in months.

What needs to land

• Coating system: typical for marine A36 is sand-blast SA 2.5 + 2-coat zinc-rich epoxy primer (~75 µm) + 2-coat polyurethane topcoat (~75 µm) = 150 µm total. ~7-10 year life in C5-M coastal.
• Drainage: bin floor low points (around weld nuts, doubler-plate seams) should drain. If standing water + salt sits in those corners, accelerated pitting.
• Wear plates: not strictly needed at CaCO₃ Mohs 3-4 against A36, but at the rear loader impact zone (where bucket teeth hit), an Ardox/Hardox 450 or AR400 wear plate bolted on is cheap insurance. Aligns with the Ardox wear-bar pattern in the drum (PRT-80xxx lifter wear bar).

Excavator dumps asymmetrically, system vibrates, ground is soft. Stability/tipping verification missing.

What's known

The full machine is feeder + drum + skid plate. CG and tipping moment are not analysed in the current feeder spec. Drum is 4 m × 1100 mm; feeder is 2.5 m × 1.65 m. Combined mass is ~2-3 t. Footprint depends on the connecting skid plate (item 8).

The case

Worst tipping case: 1-t loader dump on the rear-far-corner of the feeder bin while the machine is at 3° tilt feed-end up. The load contributes ~5 kN·m of overturning moment at the corner-support reaction. Restoring moment depends on machine weight × distance from corner support to CG.

Springs, motors, eccentric weights, pinch points, and cables require guards/protection.

Status

Tier 1 Bible §13.2 mandates feeder-drive guarding (sheet-metal enclosure, tool-only removal, Spanish warning labels, LOTO-compatible). spec.py:331: "motor_hatch: deferred to V2.1 — proper §13.2 enclosure designed once demo unit operational behavior is observed." The reasoning was "single-plate placeholder gave a false sense of compliance" — the design choice was to omit a fake guard rather than ship one.

This is a defensible posture for a demo unit operated by trained Pedro/Marco staff in a controlled fab-shop test, but it is not defensible for the 2026-06-09 hotels presentation if anyone outside the project team approaches the machine.

Have you considered if noise constraints at the hotels do not allow noisy vibrations?

Status

Bible mentions "abnormal noise" only as a commissioning fault flag, never as an upstream design constraint. Hotel acoustic limits (typical Mexican Norma NOM-081-SEMARNAT-1994: ~65 dBA daytime, 55 dBA night at property line) are tighter than industrial vibrating screens typically deliver (~85-95 dBA at 1 m).

What this changes

If hotels run the machine in back-of-house during late night for the morning beach prep (the typical sargassum-clearing window is 04:00-07:00 before guests are out), 55 dBA at the property line is a real constraint. Possible mitigations: enclosure (10-15 dBA reduction), distance (6 dB per doubling), operating-hours window, lower drive frequency (~18 Hz instead of 25 Hz, though feed performance drops).

In the assembled drawing a labeling is missing. For example I cannot find out where placa base PRT-30044 finally should go. I also cannot find that in your 3D model. Should look like that in the picture. — Andru, 22:36

What's there

Tier 1 The part exists with a clean spec at part_specs.py:113: "PRT-30044 — Placa base 180×180×10, A36 10 mm, 4 pcs, soldar al pie de cada pata, centrada, marcar centro con punzón (perno M20 jack-screw), tolerancia planitud ±1 mm, imprimación epoxi anti-corrosivo."

What's wrong

Tier 1 A second source contradicts it. wi_pdf/wi_parts_data.py:36 references "PRT-30010-P · Placa base 200×120×10mm A36 · qty 4", and step 43 reads "Soldar las 4 patas en sus respectivas placas base (200×120×10)." So the despiece part list and the WI step list disagree — 180×180×10 (PRT-30044) vs 200×120×10 (PRT-30010-P). Same role, different dimensions, different part numbers. Andru's "I cannot find PRT-30044" is correctly identifying that the part-number-as-listed-in-BOM doesn't show up in the assembly callout because the assembly callout uses the other naming.

Both files were modified in the same time window; one rename never propagated.

also weld beads are missing. most of them Pedro knows by himself best, but maybe not all of them. And if he has to assume himself he might end up welding sth that should be welded. — Andru, 22:38

What's there

Tier 1 wi_parts_data.py tags individual steps with prose welds — "weld": "soldadura continua" at lines 43, 50, 54, 60, 107, 111, 112, 116, 121, 161, 166, 178, 183, 188, 233, 235. So the WI body tells Pedro which joints get welded and that they should be continuous.

What's missing

AWS-style weld symbols on the despiece drawings. Not specified anywhere:

• Weld type (fillet vs groove vs plug)
• Weld leg / size (5 mm fillet? 6 mm? 8 mm?)
• One side vs both sides
• Continuous vs intermittent (e.g. 50/100 stitch)
• Field weld vs shop weld

For a vibration-loaded structure this matters. Andi's item 11 already named weld fatigue at the motor crossmembers as a real failure mode. If Pedro picks 4 mm fillet where 6 mm is needed, the structure under-passes design life. Andru is calling out the same risk from a different angle.

the plates with holes will all be laser parts? they lack measure for hole location. (even if it's a laser part that should be included) — Andru, 22:41

What's there

Tier 1 matplotlib_dim_renders.py exposes a holes parameter (line 237: "holes: list of {x, y, d, label?} drilled-hole specs in top-view coords") and a cutouts parameter for slots. So the renderer can draw hole positions. But the spec annotations in part_specs.py for several plates list counts ("12 × M12 holes") without listing X-Y coordinates per hole — meaning the despiece pages render the holes as circles but don't always dimension them.

Why Andru is right even for laser parts

The DXF lock the hole positions geometrically — Pedro's laser shop can cut from the DXF without dims. But:

• QC verification needs the dims on the drawing — Pedro's shop measures the cut piece against the despiece, not against the DXF.
• If the DXF is regenerated mid-project (we've done this multiple times), a missing on-drawing dim means there's no cross-check between DXF and drawing.
• Standard fab-shop practice is full dims on the drawing regardless of cut method.

Is there a decision note for me or a meeting transcript from the meeting where it was decided to use vibrations finally as feeding solution? I would like to know the reasons and test results.

No single decision note exists — the decision was made in person between Apr 20 (Marco conversation) and Apr 22 (bible v3.0 cut). The trail of evidence is reconstructable from primary sources:

1. Rejection of the "guider ramp" concept for clogging risk. field_transcripts/20260420_marco_meeting.md line 632: "harder that it gets clogged. Just like a guider, not a funnel." Wet sargassum is fibrous; converging tapers full of fibrous material are the worst-case clog scenario. Bible §11.3-11.4 lands on Plan D explicitly because it "avoids the most dangerous feeder geometry: a converging taper full of fibrous wet sargassum."
2. Mar 6 toy-car test was the first vibration probe. Per next_test_spec.md: "The toy-car test (Mar 6) answered these only partially — vibration was too weak to move wet material." That was sub-watt; failure mode was under-power, not "vibration doesn't work."
3. Apr 21 next-test-spec scaled the experiment to industrial vibration. OLI MVE motor at 3.0-3.5 kN, two rig sizes (800×1200 and 1000×1500), 3 hole sizes × 3 incline angles × 3 frequencies = 54 data points per rig. This is the load-bearing piece of evidence: it converts "vibration didn't work at toy scale" into a proper validation matrix at industrial scale.
4. Plan-D feeder geometry committed. feeder_plan_d.py is the parametric source; the bible v3.0 (Apr 22, supersedes v2.1) names Plan D as the canonical V2 feeder.

Direct answer to Andi: the decision was in person; the rationale chain is (a) wet fibrous sargassum clogs converging tapers, (b) Mar 6 vibration probe failed under-power, (c) Apr 21 industrial-scale rig design closes the validation loop, (d) Plan D = canonical. Test results from the Apr 21 rig builds (Pedro / Marco) close the loop empirically once those tests run.

PRT-30043: this solera is 76.2 mm thick. true or mistake?

Tier 1 Confirmed: cadquery_part.py:254 builds it as box_xyz(short_w=150, width, tube=76.2, ...) — solid 150 × 76.2 mm flat bar (3" × 6" A36, imperial-derived). The lower spring cup bolts into PCD 115 mm pattern through the solera; thinner section can't carry the bolt circle.

Forks

• Fork A — keep solid 150 × 76.2 mm flat bar. Pro: simplest, one cut. Con: ~208 kg of A36/unit.
• Fork B — PTR (rectangular tube) + welded reinforcement blocks at bolt-circle locations. Pro: ~60% mass reduction. Con: more welding + QC.
• Fork C — channel section (C profile). Pro: thin, strong, standard stock. Con: bolt-circle through web requires backing plate at each spring.

Recommendation deferred to Pablo's call.

Pre-fab blockers (do before Pedro cuts)

1. Item 2 — Pablo decides: bin floor solid or perforated. If perforated, lock hole size + pitch + open area + edge distances.
2. Item 3 — local plate-stress hand-calc for one-sided 1-t dump. Decide rib pitch / floor thickness before fab.
3. Item 4 — re-do slug analysis under one-sided dumping + drop-height impulse. Adjust springs or operating procedure.
4. Item 6 — spec bump stops + lateral guides; add to fab pack as a new BOM block.
5. Item 11 — weld-fatigue calc at motor crossmembers. If FAT > 36, change tube spec or bolt the motor mount.

Pre-Annex-B (2026-05-31)

1. Item 8 — open Machine-Skid WI under v2-machine-combined.
2. Item 12 — fastener spec block (Nord-Lock + Loctite + re-torque schedule + inspection access).
3. Item 13 — coating spec block (system, DFT, surface prep, drainage holes, wear-plate scope).
4. Item 14 — tipping calc once skid footprint is locked. Procedural dump-zone envelope on cover.

Pre-hotel-demo (2026-06-09)

1. Item 15 — pull §13.2 guarding forward, or formally restrict demo approach to project staff.
2. Item 16 — confirm hotel dB limits + permitted hours; budget enclosure if < 75 dBA.

Documentation tightening (low-cost, high-clarity)

1. Item 1 — bible §11 callout to Q17 + Q17 contract on fab-pack cover.
2. Item 5 — include spring-stack engineering block in fab-pack body. Close vendor-datasheet validation prompt.
3. Item 7 — lock motor model + datasheet block + vibration-direction arrow + sync mode.
4. Item 9 — render PRT-70015 spherical pivot in the WI assembly view.
5. Item 10 — section view of sand path through deflector + 150-200 mm Y-edge extension.
6. Item 17 (R2) — reconcile PRT-30044 ↔ PRT-30010-P naming + show placa base in the assembly view + confirm in SW model.
7. Item 18 (R2) — AWS weld-symbol callouts on the despiece for the vibrating sub-frame + motor cradle joints.
8. Item 19 (R2) — full hole-location dims (X, Y, Ø) on every plate's despiece, even laser parts.

Open forks (Pablo's call)

1. PRT-30043 — Fork A (solid solera) / Fork B (PTR + reinforcement) / Fork C (channel).