DXFX: Extending Model-Based Definition to 2D CAD for Plug-and-Produce Automation

Intelligent industrial automation increasingly depends on machine-interpretable information to reduce engineering effort, accelerate reconfiguration, and preserve continuity across the product lifecycle. In reconfigurable manufacturing, new products or variants require expert retuning, manual programming, and ad hoc integration of inspection logic across quality control, including Vision Inspection Systems (VIS). Model-Based Definition (MBD) advances this goal by embedding product and process semantics in CAD models; however, although MBD adoption has largely matured in 3D (e.g., STEP AP242), 2D CAD workflows remain underserved despite their continued relevance in subcontracted manufacturing and documentation. This work proposes DXF eXtended (DXFX), a lightweight format that brings MBD principles to 2D CAD and supports Plug-and-Produce (P&P) reconfiguration of inspection pipelines. DXFX augments standard DXF geometry with layered, stand-off semantics encoded in JSON: annotations persistently reference geometric targets (primitives, loops, regions) without modifying the original DXF, preserving interoperability while enabling downstream automation. Feasibility is demonstrated by formalizing and instantiating a machine-readable dataset of common punched sheet-metal defects. The enriched model drives a P&P 2D VIS that uses DXFX to configure inspection tasks and resources and writes inspection outcomes back into the same container. Results show that inspection routines can be defined and reassigned to geometric entities within minutes and executed without manual coding, enabling rapid adaptation in high-variability production where frequent reconfiguration often makes automation cost-prohibitive.