Every feature and tolerance, read from one file.
ARCNM turns a STEP file and a 2D drawing into structured, typed part data — features, threads, wall thickness, PMI and GD&T — with every value bound to the geometry it came from. A deterministic B-rep walker reads the 3D; vision-language models read the drawing; the two are fused so a tolerance attaches to the face it constrains. Reuse it for MDM, ERP and PLM, or feed it straight into process simulation and should-cost.
Two engines, one typed result
A deterministic OCC B-rep walker recognises holes, threads, pockets, bores, bends and wall thickness from the 3D solid, with stable feature IDs. Vision-language models lift dimensions, tolerances, datums and surface finish from the 2D drawing, validated against deterministic ISO grammars (ISO 286, 14405, 1101, 5459, 7200, 16792, 21920) — so machine learning never silently drives a number.
Automatic feature recognition
Recognised features come back typed and versioned: holes, threaded holes, pockets, counterbores, bores, turning grooves, bends, cutouts, notches, ribs and screw bosses, each with its dimensions, tolerances and surface finish bound in.
A first-class, reusable output
The structured part data is yours to keep — backfill ERP/PLM master data without manual keying, find similar parts, or feed the same extraction into the process and cost layers. One extraction, the whole pipeline.