Automated Construction of Dynamic Master Logic Models as Knowledge Graphs Using LLMs
A novel framework has been developed to automate the creation of Dynamic Master Logic (DML) models from system descriptions, transforming them into Knowledge Graphs (KG-DML) through Retrieval-Augmented Generation and Large Language Models. This research, detailed in arXiv (ID: 2608.12304), builds on earlier studies focused on smaller systems, expanding to much larger and intricate systems. The construction method traverses the DML hierarchy with specific retrieval, maintaining functional dependencies and clear logical connections. The resulting KG-DML facilitates diagnostic reasoning, safety evaluations, and upward failure propagation. This innovative approach overcomes the scalability challenges faced by conventional DML construction, which depends heavily on expert analysis of technical documents, positioning the framework as a vital tool for diagnosing complex systems.
Key facts
- Framework automates DML model construction from system descriptions.
- Uses Retrieval-Augmented Generation and Large Language Models.
- Represents DML models as Knowledge Graphs (KG-DML).
- Extends prior work to larger and more complex systems.
- Preserves functional dependencies and logical relationships.
- Supports diagnostic reasoning, safety assessment, and upward failure propagation.
- Addresses scalability limitations of expert-based DML construction.
- Published on arXiv with ID 2608.12304.
Entities
Institutions
- arXiv