Model-based definition for manufacturing systems reconfiguration

In computer-integrated manufacturing, Model-Based Definition (MBD) extends three-dimensional (3D) Computer-Aided Design (CAD) models beyond geometric representation by embedding product and process semantics that support downstream automation. Within Reconfigurable Manufacturing Systems (RMS), productspecific software modifications, task reprogramming, and the configuration of heterogeneous resources still impose substantial reconfiguration effort, particularly in high-mix, low-volume production. This paper investigates whether an MBD dataset can serve as the product-specific semantic input for no-code logical reconfiguration after baseline workcell deployment. The proposed approach is based on eXtended CAD (X-CAD), the CAD-oriented branch of the eXtended Formats (X-Formats) framework, which enriches CAD carriers with structured semantics. Its central contribution is the operational use of X-CAD-based semantics, encoded here in the tessellated STL eXtended (STLX) format, to instantiate task templates, bind and configure resources, activate computational routines and control-flow rules, and write selected execution feedback to the product model. The approach was validated on a multi-resource workcell that integrated an industrial robot, a linear axis, a lighting system, a vision system, and a drilling head to handle, drill, and inspect small, highly variable jewelry components. Feasible plans were generated for all seven product variants using their STLX datasets as the sole product-specific semantic input. Average annotation time was below 3 min per product, and no product-specific software or robotprogram modifications were required after baseline deployment. The results show that X-CAD-based MBD can shift product-specific reconfiguration effort from direct software modification to structured semantic annotation, thereby supporting product-centric, no-code reconfiguration across heterogeneous resources and processes in high-mix, low-volume manufacturing.