Intent-Driven Smart Manufacturing Integrating Knowledge Graphs and Large Language Models
arXiv:2602.12419v1 Announce Type: new Abstract: The increasing complexity of smart manufacturing environments demands interfaces that can translate high-level human intents into machine-executable actions. This paper presents a unified framework that integrates instruction-tuned Large Language Models (LLMs) with ontology-aligned Knowledge Graphs (KGs) to enable intent-driven interaction in Manufacturing-as-a-Service (MaaS) ecosystems. We fine-tune Mistral-7B-Instruct-V02 on a domain-specific dataset, enabling the translation of natural language intents into structured JSON requirement models. These models are semantically mapped to a Neo4j-based knowledge graph grounded in the ISA-95 standard, ensuring operational alignment with manufacturing processes, resources, and constraints. Our experimental results demonstrate significant performance gains over zero-shot and 3-shots baselines, achieving 89.33\% exact match accuracy and 97.27\% overall accuracy. This work lays the foundation for
arXiv:2602.12419v1 Announce Type: new Abstract: The increasing complexity of smart manufacturing environments demands interfaces that can translate high-level human intents into machine-executable actions. This paper presents a unified framework that integrates instruction-tuned Large Language Models (LLMs) with ontology-aligned Knowledge Graphs (KGs) to enable intent-driven interaction in Manufacturing-as-a-Service (MaaS) ecosystems. We fine-tune Mistral-7B-Instruct-V02 on a domain-specific dataset, enabling the translation of natural language intents into structured JSON requirement models. These models are semantically mapped to a Neo4j-based knowledge graph grounded in the ISA-95 standard, ensuring operational alignment with manufacturing processes, resources, and constraints. Our experimental results demonstrate significant performance gains over zero-shot and 3-shots baselines, achieving 89.33\% exact match accuracy and 97.27\% overall accuracy. This work lays the foundation for scalable, explainable, and adaptive human-machine
Executive Summary
The article presents a novel framework that integrates instruction-tuned Large Language Models (LLMs) with ontology-aligned Knowledge Graphs (KGs) to facilitate intent-driven interactions in smart manufacturing environments. By fine-tuning the Mistral-7B-Instruct-V02 model on a domain-specific dataset, the authors enable the translation of natural language intents into structured JSON requirement models. These models are then semantically mapped to a Neo4j-based knowledge graph grounded in the ISA-95 standard, ensuring operational alignment with manufacturing processes, resources, and constraints. The experimental results demonstrate significant performance gains over baseline models, achieving high accuracy in intent translation. This work lays the foundation for scalable, explainable, and adaptive human-machine interfaces in Manufacturing-as-a-Service (MaaS) ecosystems.
Key Points
- ▸ Integration of LLMs and KGs for intent-driven smart manufacturing
- ▸ Fine-tuning of Mistral-7B-Instruct-V02 on domain-specific datasets
- ▸ Semantic mapping to Neo4j-based knowledge graph aligned with ISA-95 standard
- ▸ Achievement of high accuracy in intent translation
- ▸ Potential for scalable, explainable, and adaptive human-machine interfaces
Merits
Innovative Framework
The integration of LLMs and KGs represents a significant advancement in the field of smart manufacturing, enabling more intuitive and efficient human-machine interactions.
High Accuracy
The framework achieves impressive accuracy in translating natural language intents into structured requirement models, demonstrating its effectiveness in real-world applications.
Standard Alignment
The use of the ISA-95 standard ensures that the framework is operationally aligned with existing manufacturing processes, resources, and constraints, enhancing its practical applicability.
Demerits
Limited Dataset
The fine-tuning of the LLM is based on a domain-specific dataset, which may limit the generalizability of the framework to other manufacturing contexts.
Complexity
The integration of multiple technologies and standards adds complexity to the framework, which may pose challenges for implementation and maintenance.
Scalability
While the framework is designed to be scalable, the practical scalability in large-scale manufacturing environments remains to be thoroughly tested.
Expert Commentary
The article presents a compelling framework that addresses a critical need in smart manufacturing: the translation of high-level human intents into machine-executable actions. The integration of instruction-tuned LLMs with ontology-aligned KGs is a significant innovation, as it enables more intuitive and efficient human-machine interactions. The fine-tuning of the Mistral-7B-Instruct-V02 model on a domain-specific dataset demonstrates the potential for high accuracy in intent translation, which is crucial for real-world applications. The semantic mapping to a Neo4j-based knowledge graph aligned with the ISA-95 standard ensures operational alignment with existing manufacturing processes, enhancing the framework's practical applicability. However, the complexity of the framework and the limited dataset used for fine-tuning may pose challenges for implementation and scalability. Future research should focus on expanding the dataset to improve generalizability and addressing the complexity to ensure seamless integration into existing manufacturing ecosystems. Overall, this work lays a strong foundation for scalable, explainable, and adaptive human-machine interfaces in smart manufacturing, with significant implications for both practical applications and policy development.
Recommendations
- ✓ Expand the domain-specific dataset to improve the generalizability of the framework to diverse manufacturing contexts.
- ✓ Conduct thorough testing of the framework in large-scale manufacturing environments to validate its scalability and robustness.