ENGINEERING PREVIEW — UNRELEASED

YULONG玉龍Back to the bike

Upcoming work

What is still missing

The main site tells the bike. This page is the public list of upcoming improvements on the engineering preview — a total projection of the canonical blockers, not a claim that the gaps have closed.

Rear flat-mount boss

The rear C-080 pad is fused; through-holes stay void so frame × rear_flat_mount_34 is bolt/drill clearance, not engagement. Thread form and inserts are not modelled.

  • The rear C-080 pad is fused; through-holes stay void so frame x rear_flat_mount_34 is bolt/drill clearance, not engagement. Thread form is not modelled. Inserts for the BB, ZS44 and ZS56 headset cups, and both rear dropouts are modelled as discrete bonded bodies.

UDH sweep and stop

The UDH backward rotational stop is not designed yet, and the rear derailleur body and pulley sweep are not modelled.

  • The UDH backward rotational stop is designed as frame fill in the published R14.5-19.0 stop-face band (designer protrusion, not a 60 deg sector). 60 deg minimum backward rotational clearance is proven by sweeping SRAM hanger 00-7918-089-000_REV_H about the axle.
  • The Road AXS rear derailleur body/pulley sweep is not modelled. R1 recorded the two required models as unobtainable; they were in fact found inside the UDH Rev H zip's STEP file. They remain unmodelled only because the Sheet 5 X-axis sign convention is unverified - see open_items.road_rd_keepout.

Headset cups

Headset cup envelopes are modelled as solid cylinders rather than annuli, so the steerer intersects them.

  • Headset cup envelopes remain keep-out cylinders rather than annuli, so the steerer necessarily intersects them. The manufacturing bodies are the ZS44 and ZS56 inserts. See the interference matrix rationale for those pairs.

Unreleased and not certified

This preview is not released for manufacture and is not safety certified. It is not a rideable bicycle.

  • NOT RELEASED FOR MANUFACTURE. NOT SAFETY CERTIFIED.

Visual surrogates

Bought-in components on this preview are visual surrogates placed on CAD datums. They are never interface or clearance evidence.

  • All bought-in components (wheels, tires, drivetrain, brakes, cockpit, saddle) are visual surrogates, never owned geometry: some carried from design/yulong-a45.blend unmodified, some extracted from the licensed donor assembly and refitted, one (the chain) built by the pipeline. They are placed on published CAD datums so the assembly reads correctly, but they are never clearance or interface evidence.
  • The drive chain is BUILT by the Blender stage, not appended from any library: a closed belt solved over the placed chainring, the cassette cog the derailleur is positioned over, and both jockey pulleys, then populated with links on the published 12.7 mm pitch. Its wrap angles and its clearance through the cage corridor are asserted per build, and the loop is forced to an even link count so outer and inner links alternate through the seam - but plate, roller and pin proportions are generic assumptions tagged as such, and no dimension on it is drivetrain, wrap-capacity or clearance evidence.
  • The rear derailleur, both brake calipers, the crankset with its chainring and the cassette are extracted from the licensed Rondo Ruut CF2 donor assembly (recorded in design/materials/SURROGATES.md, hash-verified at build time), de-branded by stripping every image texture from their materials, and fitted to published CAD datums by measured transforms. They are visually close bought-in stand-ins, never the specified parts and never interface evidence.
  • Recorded deviations in those donor parts, none of them hidden: the donor chainring is a 48T scaled onto the published 42T envelope, so its tooth COUNT is unchanged and its tooth pitch reads about 6 percent under the 12.7 mm the chain is built on; the donor cassette is a 12-cog stack scaled up to the declared 46T envelope; and the crank arm outer faces sit about 8 mm outboard of the published 62.5 mm W5/W6 surface, centred symmetrically rather than stretched.
  • The donor derailleur body is a cable-actuated HANGER-MOUNT derailleur while the product claims a wireless UDH Full Mount (WEB-020 / INT-004). The cable, housing, barrel adjuster and clamp are removed so no still can contradict the wireless claim, and the mount point used is the published UDH one - but the body FORM stays hanger-mount class, and the pulley tooth count is derived from the measured tip circle rather than published.
  • The baseline blend was authored with the drive side on +Y; CAD R1 places the drive side on -Y, so the surviving BASELINE surrogates are mirrored to agree with the structural master, which reverses their handedness. Handedness-sensitive detail (thread direction, brand marks) is therefore not trustworthy on those objects. The donor-extracted parts above are not affected: they are seated on the CAD drive side by measured transforms.
  • Surrogate detail level is uneven. Wheels, hubs, spokes, the cockpit and the saddle/seatpost are still the low-detail baseline set, and the rotor track is a coarse revolve. They are seated correctly but should not be read as the specified components.
  • The flat-mount ADAPTERS between structure and caliper - F20 front, FD20 rear - are dimensioned visual surrogates built by the Blender stage on published datums (SRAM GEN.0000000005232 rev V p.8 bolt radii, Shimano C-079 Ver. 1.3 mounting-face plane). The donor carries no separable bracket, so each adapter's span is solved between the MEASURED structure face and the MEASURED caliper mount face; neither standard publishes the bracket's body form, so its shape is a plausible envelope, not the specified part, and it is not CAD-master geometry or clearance evidence. It exists because the published assembly is structure -> adapter -> caliper and review 2 P1-1 found the caliper apparently bolted to nothing.
  • The caliper's lateral station is set by the measured mount face and the solved adapter standoff, not by a modelled pad centreline: SRAM Rev V and Shimano Ver. 1.3 both decline to publish the pad-centreline offset because it is internal to the caliper. The rotor-plane offset is therefore a tagged assumption at both ends.
  • YULONG dragon identity is now product-owner-approved DECAL ARTWORK, not procedural geometry. Every previous dragon fragment - ember-eye discs, gold down-tube spine, cinnabar brush strokes, gold scale discs - has been removed. The replacements are alpha-keyed transfers projected onto the CAD surface from design/materials/graphics/dragon-head-emblem.png, dragon-spine-downtube.png and the accent set (dragon-accent-claw.png, dragon-accent-whisker.png, dragon-accent-flame.png). They are zero-relief appearance only and carry no structural meaning.
  • The brake rotors are BUILT by the Blender stage, not appended from the baseline blend, because the baseline rotor was a flat annulus with six straight arms and read as a featureless disc. The built rotor uses the published 160 mm nominal diameter and rotor plane offsets, but its slot pattern, spider arm form and carrier proportions are a plausible visual surrogate and are NOT the specified Paceline part.
  • The donor component set is adopted PER PART, never wholesale. A wholesale swap was evaluated in round 4 and rejected because it would have put ~880 mm diameter donor tires on a frame dimensioned by CAD for 700x45 (the recorded design tire), misrepresenting exactly the tire/frame clearance this channel exists to show. The tire therefore stays the conformed 700x45 design casing inside the CLR-005 keep-out, and each adopted part is extracted and fitted on its own datums.
  • Visual-stage artwork is independent of the CAD structural revision. Do not treat Blender-only changes as closed CAD modelling gaps; compare cad_revision and STEP/tessellation checksums to the previous preview when judging whether structure moved.

Additional unresolved work

  • CAD A45-CAD-STARTER-R6 modelling gap: Seat-clamp collar, slot and bore relief are modelled against the 27.2 mm post / 27.4 mm bore. Clamp force and torque are not declared.
  • CAD A45-CAD-STARTER-R6 modelling gap: Heel/shoe and the front-tire steering sweep are modelled as assumption-tagged packaging envelopes (unpublished foot last and steer band, placed from published Q-factor and L5). They are keep-outs, not published rider or lock geometry.
  • CAD A45-CAD-STARTER-R6 modelling gap: The XD/XDR driver body is not modelled: SRAM publishes no driver body outer diameter or length, and xddriverbody.com carries no dimensions.
  • CAD A45-CAD-STARTER-R6 modelling gap: The cassette's cog pitch and stack width are unpublished assumptions. The stay clearance verdict is proven insensitive to them by a sweep, but the envelope itself is not a controlled dimension.
  • CAD A45-CAD-STARTER-R6 modelling gap: Bore edge radii, edge breaks, thread form, bearing-seat tolerances and anti-rotation detail are not modelled. Bond lines exist as declared 0.3 mm ASSUMPTION solids at the five insert pockets (BB, ZS44, ZS56, both rear dropouts).
  • CAD A45-CAD-STARTER-R6 modelling gap: The nine STR-003 tube meetings (BB/DT, BB/ST, BB/CS left and right, HT/DT, HT/TT, ST/TT, ST/SS left and right) carry OCC seam-fillet blends with named ASSUMPTION radii; those radii are not supplier-validated.
  • CAD A45-CAD-STARTER-R6 modelling gap: Down-tube pocket walls, hatch, lip, cover, bottle-boss stations and the internal routing corridor are packaging assumptions pending factory/ODM drawings. They are modelled so STOR-004 can fail-closed; they are not released dimensions.
  • CAD A45-CAD-STARTER-R6 modelling gap: Hollow walls are a packaging-offset second inner shell on the frame and fork, not a bladder/mandrel, laminate minimum, or mould split.
  • Structural geometry is hollow packaging envelopes (a second inner shell on frame and fork, exported as frame_inner_surface / fork_inner_surface with role inner, excluded from the GLB) plus the down-tube pocket local wall, hatch cut in frame_outer_envelope, and a separate down_tube_storage_cover component. The CAD master still carries no detailed surfacing, so the preview reads as primitive next to the hand-modeled legacy visual baseline. This is expected; Blender must not 'improve' the structural shapes.
  • Hollow walls are packaging-offset inners, not a laminate schedule; nine STR-003 meetings carry OCC seam-fillet blends with named ASSUMPTION radii; BB, ZS44, ZS56 and both rear-dropout inserts sit as discrete bonded bodies with declared 0.3 mm ASSUMPTION bond-line solids; no ply book, thread form, mould split, front-pad inserts, adhesive/prep/galvanic isolation, or REL-002 rollup; envelope wall figures are packaging assumptions, not layup prescriptions.
  • Keep-out volumes - tire, rotor, chainring, crank sweep, cassette envelope, chain corridor and rear dropout clearance zone - are tessellated as run evidence but excluded from the GLB, so the preview shows no clearance envelopes. Classification is by name pattern as well as by explicit list, so a keep-out added by a later CAD revision cannot arrive in the GLB looking like a component.
  • Failed gate 'engineering-release-verdict': engineering/tools/preflight.py release verdict (--strict gate).
  • Failed gate 'preflight-tests': engineering/tests unittest suite.

Preview facts

CAD rev
A45-CAD-STARTER-R6
Gates
2 passed / 2 failed / 0 not run
Blockers
31
STEP
cc843211ef6f
GLB
1a5f5049d183