From ink to insight: A manufacturing data space empowering transfer learning for predictive failure in industrial offset printing

Although offset printing remains the dominant high-volume print manufacturing method, the industry struggles with process optimization, waste reduction, and proactive maintenance. Current control strategies are often limited by a heavy reliance on manual expertise and fragmented, ”siloed” machine data. This paper introduces a Data Space for offset printing, establishing a framework for sovereign and secure data exchange between printing companies and service providers. In addition, the study integrates this Data Space with Transfer Learning, facilitating rapid model adaptation and cross-organizational knowledge sharing. This dual approach addresses data scarcity and scalability, contributing to the development of more resilient and automated printing ecosystems. To validate the proposed approach, an experiment was conducted in a real-world setting involving four distinct organizations.