ENTERPRISE AI ANALYSIS
Empowering Model-Based Systems Engineering with AI-Driven Information Retrieval
This research introduces an LLM-based multi-agent system leveraging GraphRAG to provide intuitive and accurate access to complex system models, overcoming challenges in information retrieval and enhancing MBSE accessibility for all stakeholders.
Key Outcomes & Performance Metrics
0
Average Accuracy (Best Model)
0
LLMs Evaluated
0
Nodes & Relationships in Model
0
Curated Questions in Dataset
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
MBSE Adoption & Challenges
Model-Based Systems Engineering (MBSE) improves traceability and consistency for complex cyber-physical systems, but its adoption faces hurdles like high initial effort, steep learning curves, and difficulty retrieving information from large models.
0.6%
of scientific papers empirically measured MBSE benefits [4]
Barriers to MBSE Implementation
Key obstacles include cultural resistance, lack of management commitment, tooling challenges, restrictive IT policies, steep learning curves, and difficulties in measuring Return on Investment (ROI) [15,16]. These human and technological factors often require significant upfront investment and early anticipation [5,17].
Generative AI in MBSE: Current Focus
While generative AI, especially Large Language Models (LLMs), shows promise for MBSE, current research primarily targets the introduction phase—assisting with model generation, requirement analysis, and automated consistency checks—rather than operational use and information retrieval from existing models [7-10,21].
Unaddressed Potential: Operational Use Phase
Despite some proofs of concept [11], significantly less attention has been given to facilitating the use of information stored in MBSE models during the operational phase. This gap is crucial to address for realizing the full potential of MBSE, allowing stakeholders and AI to access complex system data naturally.
GraphRAG for MBSE System Models
Graph Retrieval-Augmented Generation (GraphRAG) is uniquely suited for MBSE, as system models inherently possess a graph-like structure of entities and relationships. This approach overcomes the limitations of conventional RAG by enabling precise retrieval and reasoning over complex interconnections.
| GraphRAG Component | Our Approach |
|---|---|
| Indexing Method | Hybrid (Graph + Vector) |
| Retriever Type | LM-based (Multi-Agent System) |
| Retrieval Paradigm | Iterative, Adaptive (ReAct pattern) |
| Retrieval Granularity | Hybrid (All Granularities via Cypher) |
| Query Enhancement | Query Decomposition (Supervisor Agent) |
| Generator | LM (Same Model as Retriever) |
| Graph Format | Graph Language, Code-Like (Cypher, JSON) |
Methodology & Performance Evaluation
Our methodology involves converting SysMLv2 models into a knowledge graph and querying it with a hierarchical multi-agent system using a customized GraphRAG strategy. Evaluated on a Battery Electric Vehicle (BEV) architecture, the best-performing LLM achieved 93% accuracy on 100 curated questions.
Enterprise Process Flow
Phase 1: Preprocessing Pipeline (SysMLv2 → Knowledge Graph Metamodel Translation)
→
Phase 2: Multi-Agent System (GraphRAG Implementation, Customized Retrieval Strategy)
→
Phase 3: Reference Model (Battery Electric Vehicle RFLP Architecture in SysMLv2)
→
Phase 4: Evaluation (Q/A Dataset, Performance Assessment)
| Question Category | Gemini 2.5 Flash | Gemini 2.0 Flash | Gemini 2.0 Flash Lite | Llama-3.3-70B-Instruct-Turbo |
|---|---|---|---|---|
| One-hop | 96% | 94% | 88% | 94% |
| Multi-hop | 90% | 82% | 64% | 50% |
| Average | 93% | 88% | 76% | 62% |
Battery Electric Vehicle (BEV) Reference Architecture
A comprehensive BEV model developed in SysMLv2 textual notation, covering requirements (97), functions (37), logical (9), and physical (34) components. This forms a knowledge graph with 427 nodes and 505 relationships, ensuring full traceability across RFLP abstraction levels. This model served as the robust foundation for evaluating our AI-based retrieval system.
Advanced ROI Calculator
Quantify the potential impact of AI-driven MBSE on your engineering efficiency and cost savings, based on our demonstrated improvements in information retrieval.
Potential Annual Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A structured approach to integrating AI-driven MBSE information retrieval into your enterprise, based on our proven methodology.
Phase 1: SysMLv2 to Knowledge Graph Transformation
Develop a custom parser and graph schema to convert SysMLv2 models into a Neo4j knowledge graph, preserving all structural and semantic relationships.
Phase 2: Multi-Agent GraphRAG System Development
Implement a hierarchical multi-agent system with a Supervisor Agent and Graph Query Agent, leveraging an iterative and adaptive GraphRAG strategy.
Phase 3: Reference Model Application & Refinement
Develop and apply the methodology to a Battery Electric Vehicle (BEV) reference architecture, ensuring full coverage of RFLP abstraction levels.
Phase 4: Comprehensive Evaluation & Benchmarking
Evaluate the system's performance using a 100-question Q/A dataset across various LLMs, assessing accuracy, response times, and identifying areas for further optimization.
Unlock the Full Potential of Your Enterprise Engineering Data
Ready to transform your MBSE workflow with AI-driven insights? Schedule a personalized consultation to explore how our solutions can enhance your system engineering processes, improve data accessibility, and accelerate decision-making.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat