Graph‐based Operationalization of Machine Learning Applications in Manufacturing

von Tim Raffin

187 Seiten, Taschenbuch

Kurzbeschreibung des Verlags

Despite its demonstrated economic potential, the broad scaling of ML applications in manufacturing frequently fails due to heterogeneous production environments, complex lifecycle management, and a shortage of qualified personnel. Although up to 94% of manufacturing companies have conducted ML proof-of-concepts, only a fraction successfully scales these initiatives into productive deployment. This thesis develops a methodology for the scalable operationalization of ML applications in manufacturing, integrating principles from SPLE, DDD, and MDSE into a coherent framework. As its central contributions, the thesis presents a domain-specific MLOps reference architecture, a graph-based data model capturing interdependencies between manufacturing systems and ML compo-nents, and a graph-based configurator enabling automated operationalization by manufacturing experts without deep ML implementation experience. The approach was validated through a case study on deep learning-based optical inspection in hairpin stator manufacturing, demonstrating its effectiveness both quantitatively and qualitatively.

• Reihe: FAU Studien aus dem Maschinenbau
• Themen: Technologie, Ingenieurswissenschaft, Landwirtschaft, Industrieprozesse Maschinenbau und Werkstoffe Maschinenbau
• ISBN: 9783961479528
• Sprache: Englisch
• Erscheinungsdatum: 09.09.2026
• Größe: 24 x 17 cm
• Verlag: FAU University Press