Systematizing Data Preparation in Smart Manufacturing via Axiomatic Design: A Toolkit Integrating GUI and Agentic AI

Abstract Data serves as the primary driving force behind smart manufacturing, yet the transition from raw, upstream shop-floor signals to analysis-ready datasets remains a critical operational bottleneck. Often dismissed as isolated janitorial work, data preparation consumes a disproportionate amount of analytical effort due to a reliance on fragmented, ad-hoc methods that inherently lack reproducibility, traceability, and scalability. To overcome these barriers, this study proposes a systematic, requirements-driven approach grounded in the principles of Axiomatic Design. Specifically, this study cross-examines diverse manufacturing experiments, identifies recurring operational bottlenecks, and maps them to specific functional requirements. This directly drives the development of a decoupled, open-source toolkit. The toolkit is realized in two complementary interaction models: an Interactive graphical user interface for granular ground-truth validation and an agentic artificial intelligence system for automated, intent-based orchestration. Validated through a micro-drilling case study, the toolkit demonstrates how complex, high-volume workflows can be systematically executed without manual drudgery. By standardizing the upstream data preparation tasks, the toolkit democratizes access to advanced workflows across diverse industrial and academic environments. Ultimately, by transforming data preparation from a peripheral burden into a rigorous scientific discipline, this systematic upstream management establishes the essential, structured foundation required to power advanced downstream applications, such as functional digital twins and cognitive decision-making systems.