From CAD mates to constraints

The compiler combines the constraints between each pair of owned moving members, determines which relative motions remain, and proposes a supported joint.

Combine the mates before choosing a jointCaptured assembly connection: Coincident20 + Concentric8

Concentric axisAxis + seating planeAxis + locked rotationPlane + locked axis

Axis + seating plane → revolute joint: one rotation remains.

Constraint matrix → remaining motion

Each row constrains a combination of translations and rotations. The nullspace contains the motions that satisfy every row. Repeated rows do not remove another degree of freedom.

Row	tx	ty	tz	ωx	ωy	ωz

Six relative degrees of freedom, five independent constraints, one remaining motion.

For the captured hinge, the supported template binds two distinct hardpoints on this axis. Both rigid groups share those pivot nodes, leaving rotation around their connecting line. A mate does not become one beam. Ownership, hardpoint bindings and the supported mechanism contract must still pass review.

Source and alternatives

Axis and plane come from saved native mate records. The two “locked” options are hypothetical edits run through the real compiler; they are not changes to your assembly. Matrices and classifications were exported from joint_authoring.classify. Data and source hashes.

How the conversion works. NativeMateScanner.cs captures persistent native geometry and assembly-space coordinates on the CAD thread. joint_authoring.py builds a relative six-DOF constraint matrix from point/axis/plane geometry and uses its nullspace plus geometric checks to propose supported motion. Mate signatures preserve approval only while the captured constraints stay unchanged. Role ownership and joint_contract.py reject wrong-group points, off-axis hinge pairs, incompatible motion and unused physical constraints. authored_shafts.py walks a connected, unbranched chain of supported joints. No mesh bound or source body order supplies a missing hardpoint.

translating_rack.py fits the two members’ positive masses independently to their CAD mass/COM/inertia. Three parallel capped rails constrain rack rotation and transverse movement; opposed steering hydros translate it, and both tie rods keep fixed lengths. Their springs share the assembled numerical step budget. The implementation derives the hydro’s control sign by solving both front linkages with a small positive and negative rack displacement. Positive steering input must turn both wheels right in vehicle coordinates. This handles racks ahead of or behind the axle and either axis direction; contradictory wheel responses stop generation. The sign and toe gains are recorded in the steering contract. The implementation converts physical rack speed to relative hydro rate using the actuator’s rest length, preserving speed when that numerical length changes. See BeamNG’s hydro controls. The implementation applies both actuators at one midpoint node between the rack endpoints, with opposite housing supports and clearance for the full stroke. The supports share the former total requested holding stiffness and damping; they target the same physical translation with opposite length changes. A single-sided numerical support introduced transverse-to-steering coupling even in a symmetric CAD rack. A regression independently solves the exported static stiffness and requires symmetric transverse loads to leave the rack center’s axial position unchanged. All support nodes receive redistributed CAD member mass, and the assembled stiffness-and-damping step limit still applies. The three guides use this assembled budget without an additional local frequency cap. Native road-load verification remains in progress. Compact rigid housings double mounting-node priority relative to their existing CAD mass allocation, then refit the same CAD mass, COM and inertia. Other nodes retain their relative allocation priority, including torque reaction supports. All nodes remain positive and the achieved allocation is reported. This is a numerical support policy; it neither adds a fixed 2 kg to each mount nor calibrates the strength of a real fastener. Normal constraints share the full assembled step budget; deformable collision cells retain an additional local cap. The motion reference predicts the wheel linkage using the recorded native hydro target, so a drifting rack cannot validate itself by supplying its own measured position as the target. Every declared actuator must compile once and record the same physical target; missing or conflicting targets reject the recording. Earlier single-actuator isolated sweeps at ±10/±40 mm passed with peak tracking errors 0.11/0.43 mm. Those recordings do not qualify the opposed-support revision. Full-vehicle road-load acceptance of this new rack still has to run on each generated vehicle.

attachment_contract.py creates three short directional connector beams per breakable fastener, with stable release groups scoped to the child. Permanent frame restraints distribute force through the calibrated frame supports. Each panel retains its own nodes, triangles and flexbody binding after release. The boundary audit rejects undeclared beams, torsionbars, triangles or rails that could keep a panel attached, and rejects a skin bound to another part’s nodes. The harness records both deliberate cuts and force-ramp failures; independent mounts retain the panel until their own connectors fail. Named linked sets release together. Numerical stability caps and CAD mass fitting do not calibrate hardware rupture.

Supported scope. Native point, concentric/locked-concentric and compound hinge fixtures were checked in SolidWorks, including reopen identity. Companion split, source refresh and body reclassification were checked on owned native geometry. The candidate also passed a fresh 36-group companion lifecycle with ten moving members, save/reopen reference checks and clean owned SolidWorks shutdown. Member splitting copies replacement parts directly into the final revision, avoiding temporary assembly windows. Body/reference lookup caches last only for one revision update, so a later source change cannot reuse stale CAD references. The current conversion does not run a native CAD finite-motion sweep and does not convert generic mate travel/angle limits or arbitrary linkages into physical stops. Those constraints are rejected rather than discarded; use the supported template’s explicit travel/coilover stop settings and mark a CAD-only limit positioning-only after review. Rack checks used a fixed housing and analytic moments; two authored vehicle graphs passed offline mass/release checks. New vehicles remain diagnostic until their own live verification passes.