Paper Overview
Field: Machine Learning Authors: Saman Marandi, Yu-Shu Hu, Mohammad Modarres Published: 2026-08-13 arXiv: 2508.03416
What This Paper Does
Dynamic Master Logic (DML) offers a hierarchical framework for representing system behavior by linking functional goals to underlying structural elements. However, DML construction has traditionally depended on experts manually interpreting technical documentation, which limits scalability to complex systems.
This work presents a framework that automatically constructs DML models from system descriptions and represents them as knowledge graphs (KG-DML), using retrieval-augmented generation (RAG) and large language models as enabling tools. Building on earlier work with small-scale systems, the framework extends automated KG-DML construction and evaluation to larger, more complex systems.
Key Contributions
- Hierarchy-aware construction: Model building proceeds within the DML hierarchy, using goal-based retrieval that preserves functional dependencies and explicit logic relationships.
- Executable functional models: The generated KG-DML supports:
- Diagnostic reasoning
- Safety assessment
- Upward fault propagation
- Downward dependency tracing
- Multi-level validation: Evaluation covers level-specific precision and recall, logic gate consistency, and overall structural integrity.
Case Study
The framework was applied to the low-pressure coolant injection (LPCI) system of a decommissioned boiling water reactor. The results demonstrated consistent reconstruction across repeated runs, showing robustness of the automated pipeline.
Takeaway
Automated KG-DML construction can transform technical documentation into executable functional models suitable for diagnostic and reliability analysis, reducing reliance on manual expert interpretation.
--- *Auto-collected on 2026-08-14*