Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Knowledge graph enhanced retrieval-augmented generation for failure mode and effects analysis

Lukas Bahr, Christoph Wehner, Judith Wewerka, José Bittencourt, Ute Schmid, Rüdiger Daub

TL;DR

This work addresses the limitations of traditional, tabular FMEA tools by marrying a knowledge-graph (KG) representation with retrieval-augmented generation (RAG). It formalizes FMEA using set-theoretic constructs, defines a KG schema to capture failure modes, effects, causes, and measures, and develops a DFS-based embedding algorithm to create vector representations from the KG. The proposed KG-RAG framework enables both graph-query and vector-based retrieval to support analytical and semantic QA on FMEA data, with a user study showing improved correctness, usability, relevance, and completeness over Excel and enhanced context recall/precision. The approach advances practical, interpretable risk analysis in manufacturing and points to broader AI-assisted risk management applications across quality domains.

Abstract

Failure mode and effects analysis (FMEA) is an essential tool for mitigating potential failures, particularly during the ramp-up phases of new products. However, its effectiveness is often limited by the reasoning capabilities of the FMEA tools, which are usually tabular structured. Meanwhile, large language models (LLMs) offer novel prospects for advanced natural language processing tasks. However, LLMs face challenges in tasks that require factual knowledge, a gap that retrieval-augmented generation (RAG) approaches aim to fill. RAG retrieves information from a non-parametric data store and uses a language model to generate responses. Building on this concept, we propose to enhance the non-parametric data store with a knowledge graph (KG). By integrating a KG into the RAG framework, we aim to leverage analytical and semantic question-answering capabilities for FMEA data. This paper contributes by presenting set-theoretic standardization and a schema for FMEA data, an algorithm for creating vector embeddings from the FMEA-KG, and a KG-enhanced RAG framework. Our approach is validated through a user experience design study, and we measure the precision and performance of the context retrieval recall.

Knowledge graph enhanced retrieval-augmented generation for failure mode and effects analysis

TL;DR

This work addresses the limitations of traditional, tabular FMEA tools by marrying a knowledge-graph (KG) representation with retrieval-augmented generation (RAG). It formalizes FMEA using set-theoretic constructs, defines a KG schema to capture failure modes, effects, causes, and measures, and develops a DFS-based embedding algorithm to create vector representations from the KG. The proposed KG-RAG framework enables both graph-query and vector-based retrieval to support analytical and semantic QA on FMEA data, with a user study showing improved correctness, usability, relevance, and completeness over Excel and enhanced context recall/precision. The approach advances practical, interpretable risk analysis in manufacturing and points to broader AI-assisted risk management applications across quality domains.

Abstract

Failure mode and effects analysis (FMEA) is an essential tool for mitigating potential failures, particularly during the ramp-up phases of new products. However, its effectiveness is often limited by the reasoning capabilities of the FMEA tools, which are usually tabular structured. Meanwhile, large language models (LLMs) offer novel prospects for advanced natural language processing tasks. However, LLMs face challenges in tasks that require factual knowledge, a gap that retrieval-augmented generation (RAG) approaches aim to fill. RAG retrieves information from a non-parametric data store and uses a language model to generate responses. Building on this concept, we propose to enhance the non-parametric data store with a knowledge graph (KG). By integrating a KG into the RAG framework, we aim to leverage analytical and semantic question-answering capabilities for FMEA data. This paper contributes by presenting set-theoretic standardization and a schema for FMEA data, an algorithm for creating vector embeddings from the FMEA-KG, and a KG-enhanced RAG framework. Our approach is validated through a user experience design study, and we measure the precision and performance of the context retrieval recall.

Paper Structure

This paper contains 13 sections, 3 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. QA for FMEA
  4. Set-theoretic formalization of FMEA
  5. Modeling FMEA data within a KG
  6. KG-RAG framework
  7. Experiments
  8. Implementation details
  9. Study methodology
  10. Results of the study
  11. Evaluation of the context retrieval
  12. Discussion of the results
  13. Conclusion

Figures (3)

  • Figure 1: Screen capture of the FMEA chatbot. The user interface is kept simple. Clicking on context reveals the information retrieved from the FMEA-KG.
  • Figure 2: Overview of the proposed schema for the FMEA, including the symbols, literals, and relations. The nodes ($n_1 - n_5$) represent a failure mode and illustrate how depth-first search is conducted on the tree-like graph structure.
  • Figure 3: Sketch of the proposed KG-RAG framework. The information is retrieved either by utilizing the graph query language of the KG or by using vector search. The framework employs an LLM to generate the query and to serve the result.