This study proposes a human-in-the-loop framework that integrates operator observations into a large language model (LLM) to control process parameters for defect handling in fused deposition modeling (FDM) 3D printing. Fully autonomous LLM-based control handles ambiguous sensor data poorly and cannot detect abnormal conditions that lie beyond the installed sensors. Operator observations may compensate for these limitations, but their actual impact on LLM decision-making has not been sufficiently validated. We therefore implemented the proposed framework and defined experimental scenarios involving erroneous parameter injection and environmental disturbances. The framework was evaluated in terms of LLM response quality and print quality. The LLM achieved over 80% response quality on the defined evaluation metrics and generated appropriate parameter adjustments, improving print quality by more than 55% on average. Comparative experiments further revealed that, without operator observations, the LLM sometimes failed to recognize defects. These findings demonstrate the effectiveness of human–LLM collaboration and provide a practical foundation for intelligent FDM process control.

As smart factories evolve, maintenance manuals need to be transformed from static documents into machine-readable and reusable digital assets. However, many legacy manuals are still in unstructured formats, such as Hangul word-processor files, which complicates their updating, reusability, and adaptability to changing product configurations. This paper presents a framework for converting these legacy manuals into S1000D-based documents. It combines style-based hierarchy extraction with rule-guided multi-step transformation using a local large language model (LLM). First, the style information within the Korean documents is analyzed to identify the hierarchical structure of the manual and extract content at various document levels. Next, this extracted content is converted into S1000D XML modules through the local LLM, utilizing category-specific rule files, XML tag definitions, and example templates. To enhance structural consistency and minimize errors, different prompts and rule sets are applied based on the document hierarchy level.A case study involving a maintenance manual for a high-angle limit switch module demonstrates that the proposed method can maintain document structure while generating reusable S1000D-style outputs from legacy technical documents. This approach lays a practical foundation for creating continuously updatable and context-reconfigurable maintenance guidance in smart manufacturing environments.

This paper examines the role of generative AI and large language models (LLMs) in advancing intelligent manufacturing as we transition from Industry 4.0 to Industry 5.0. We begin by analyzing the current limitations of rule-based and manufacturing data systems in facilitating flexible, human-centric production. Next, we categorize LLM utilization strategies into three methodological axes: fine-tuning domain-specific models, employing general-purpose models through prompt engineering, and utilizing retrieval-augmented generation (RAG), which includes multimodal RAG that integrates sensor and text data. For each strategy, we present representative case studies across key application areas such as asset management, maintenance intelligence, quality control, process optimization, and knowledge- and document-centric support systems. Concurrently, we explore how information modeling and ontology-based knowledge graphs can be integrated with LLMs to enhance structured manufacturing semantics, improve source traceability, and minimize hallucinations. Finally, we summarize the advantages and limitations of each approach and propose future research directions for human-centric manufacturing, including the development of trustworthy LLM pipelines, standardized data schemas, and closer integration between digital twins and LLM-based decision support systems.