Integrating artificial intelligence and digital twins into a data architecture for small-batch manufacturing

Abstract

This study explores the adoption of Artificial Intelligence (AI) and Digital Twins (DT) in small-batch production enabled by Additive Manufacturing (AM), with a particular focus on Small and Medium-sized Enterprises (SMEs), where limited data, fragmented toolchains, and resource and skills constraints limit effective implementation. A structured literature review based on bibliometric and content analyses following the PRISMA protocol synthesizes recent research at the intersection of AI, DT, and AM and enables the identification of design requirements. Guided by a Design Science Research methodology, the study proposes a layered, data-centric architecture that integrates a Digital Process Twin with real-time monitoring and control, bidirectional communication, and intelligence integration across two manufacturing sites. Three use cases demonstrate progressive levels of fault management, expanding the scope from manual operator intervention upon error detection to automated machine calibration for root-cause elimination and ultimately to fully autonomous closed-loop control for real-time process adjustments. The work addresses the lack of holistic and reusable integration in small-batch production and clarifies key implementation concerns, including time synchronization, latency, and human oversight. The contributions are twofold: (i) an evidence-based synthesis of technologies and themes for AI and DT supporting small-batch manufacturing, and (ii) a generalizable architecture and related use cases that aim to enable more individualized, higher-quality production in small batches.