Proceedings of the XMO Industrial Seminar 2026: Excellence in Manufacturing and Operations
Keywords
Manufacturing systems, generative AI, decision making
Tracks
DIGITAL MANUFACTURING
DOI
10.5703/1288284318690
Abstract
Manufacturing systems increasingly demand real-time, multi-objective, and human-aligned decision-making that existing approaches cannot provide. In this study, we introduce the Generative Manufacturing System (GMS), a paradigm that reframes manufacturing decisions as samples from a learned conditional distribution, replacing inference-time search in the existing optimization-based approaches. Two instantiation examples are presented in manufacturing decision-making, including layout design and production scheduling, using domain-adapted diffusion models conditioned on operational constraints and human preferences. Experiments demonstrate 100% feasibility, near-zero objective matching error, and 25x decision making speed gain, establishing generative AI as a scalable, efficient, and human-centric decision-making engine for modern manufacturing systems.
Generative AI for decision making in the manufacturing systems
Manufacturing systems increasingly demand real-time, multi-objective, and human-aligned decision-making that existing approaches cannot provide. In this study, we introduce the Generative Manufacturing System (GMS), a paradigm that reframes manufacturing decisions as samples from a learned conditional distribution, replacing inference-time search in the existing optimization-based approaches. Two instantiation examples are presented in manufacturing decision-making, including layout design and production scheduling, using domain-adapted diffusion models conditioned on operational constraints and human preferences. Experiments demonstrate 100% feasibility, near-zero objective matching error, and 25x decision making speed gain, establishing generative AI as a scalable, efficient, and human-centric decision-making engine for modern manufacturing systems.