Context Fusion Agent in Design-to-Plan Workflow

This paper presents the Context Fusion Agent, a crucial component in enhancing design-to-plan workflows in manufacturing by bridging the semantic gap between design representations and manufacturing reasoning. The objective of this study is to integrate outputs from a Feature Extraction Agent, which identifies 3D CAD features and 2D drawing specifications, into a unified feature representation that supports actionable manufacturing processes. Through a two-stage methodology involving semantic interpretation and hybrid matching, each 2D annotation is enriched with manufacturing context to facilitate accurate correlation with corresponding 3D CAD features. Results indicate that by linking dimensional, tolerance, and surface finish requirements from drawings directly to specific CAD features, the framework allows for improved knowledge retrieval, efficient process sequencing, and selection of appropriate tools. The enriched 3D feature representation significantly enhances downstream analytical capabilities, enabling the development of valid manufacturing rules and operational instructions. This approach not only addresses the ambiguities inherent in dimensional values but also incorporates expert user feedback to refine the mappings, leading to more accurate manufacturing outcomes.