Characterising the creation of digital twins in manufacturing processes from the functional perspective

This paper proposes a functional architecture for the creation of digital twins (DTs) in manufacturing processes. The architecture comprises cross-layer and cross-disciplinary functional components (FCs) involved in DT creation, aiming to address the lack of an explicit functional basis for systematic requirement allocation and early-stage cost estimation. A hybrid methodology that combines top-down structuring with bottom-up enrichment is implemented by integrating standards, expert knowledge, and insights extracted from existing DT studies. Specifically, the functional entities (FEs) defined in the ISO 23247 DT architecture for manufacturing are decomposed to construct a seed taxonomy through a top-down process. In parallel, bottom-up enrichment is achieved through literature-driven text analysis and concept extraction. A corpus of DT-related literature is analysed to identify additional FCs beyond those specified in the seed taxonomy. Human experts are involved to refine and consolidate the generated outputs. In total, 401 documents are analysed, yielding 13,326 and 13,291 FC instances using GPT-5 and Gemini 2.5 Pro, respectively. The resulting functional architecture comprises 107 FCs across 23 FEs, along with a strategic guideline for its application in DT creation. A robotic drilling case study is used to illustrate the application of the architecture in structuring early-stage DT creation. In an exploratory within-case comparison, two engineers produced DT designs with weighted requirement satisfaction of 95.6% and 91.1% using the proposed architecture, compared with 68.9% and 67.8% under ISO 23247 and 67.8% and 57.8% under a unified step-by-step guideline.