3D Spare Parts Catalogs in Manufacturing

This thesis investigates the implementation of 3D visualization technologies for industrial spare parts catalogs, addressing technical challenges of legacy data integration and enterprise system connectivity. The research question asks: How can manufacturing companies implement 3D visualization technologies for industrial spare parts catalogs while addressing the specific technical challenges of legacy data integration and enterprise system connectivity? The investigation follows Design Science Research methodology, developing and evaluating a functional prototype system at Signifikant Svenska AB. The research examines three sub-questions addressing architectural patterns for deploying 3D visualization within existing ERP systems (RQ1), technical challenges in converting legacy CAD data to web-optimized formats (RQ2), and system performance in addressing limitations of traditional 2D documentation (RQ3). The research employed automated conversion pipeline transforming legacy STEP CAD files to web-optimized glTF format, hybrid YOLOv8+EasyOCR system for extracting metadata from legacy 2D technical drawings, browser-based 3D rendering using Three.js and WebGL tested across desktop, tablet, and smartphone devices, and performance evaluation measuring conversion success rates, OCR precision, loading times, and frame rates. Results demonstrate technical implementation feasibility: 100\% conversion success rate, 99.3\% OCR precision, acceptable rendering performance (31-58 FPS) across device categories, and load times under 5 seconds over simulated 3G networks. Three architectural patterns emerged: stateless client design, metadata embedding within geometry files, and event-driven bidirectional synchronization. Critical limitations: evaluation focused on technical implementation feasibility only (not business value), limited samples preventing broad generalization, conversion success rates specific to STEP archives, and claimed operational benefits were not empirically measured. The thesis contributes quantified conversion success boundaries, browser-specific performance characterization, two-stage OCR architecture advantages, and reusable architectural patterns for industrial ERP-3D integration.