About this Abstract |
| Meeting |
2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026)
|
| Symposium
|
2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026)
|
| Presentation Title |
Human-AI teaming for information model DEVELOPMENT - A benchmark study for AM-MES integration |
| Author(s) |
Sebastien Philomin, Yan Lu |
| On-Site Speaker (Planned) |
Yan Lu |
| Abstract Scope |
Developing standardized information models for Industry 4.0 interoperability remains a time-intensive, expert-driven process. Large Language Models (LLMs) suggest potential acceleration of model engineering tasks, yet their ability to generate structurally valid and semantically coherent industrial information models under rigorous constraints remains unverified. This study investigates whether human–AI teaming can accelerate Additive Manufacturing (AM) to Manufacturing Execution System (MES) integration. We propose a benchmark evaluating three strategies: (1) traditional human-only modeling; (2) direct LLM-based model generation; and (3) a constrained human–AI collaborative approach, where experts define standards-based constraints (e.g., OPC UA) and AI operates solely as a semantic reasoning engine mapping heterogeneous AM data. Experimental results show unconstrained LLM generation fails consistently due to syntax hallucinations and inherent context window limitations. In contrast, the constrained human–AI approach achieves full structural correctness and superior semantic alignment while reducing development effort, introducing benchmark metrics to evaluate AI-driven integration. |
| Proceedings Inclusion? |
Undecided |