1. BaSaMa & HiWi
    2.  

LLM-Based Information Extraction for Populating a Disassembly Knowledge Graph

Institute
Institut für Werkzeugmaschinen und Betriebswissenschaften (TUM-ED)
Type
Bachelor's Thesis / Semester Thesis / Master's Thesis /
Content
experimental / theoretical / constructive /  
Description

 

Disassembly process planning requires structured knowledge that captures both the feasible disassembly sequences of a product and the operational information needed for execution. Recent work has produced a knowledge graph schema that formally integrates AND/OR decomposition logic with operational process semantics within a unified, queryable representation. The schema follows a directed layered architecture: the structural layer is populated automatically from CAD data, while the operational layer (tasks, skills, tools, agents) is currently populated through manual entry via structured tables.

In industrial practice, however, the required operational information is dispersed across unstructured technical documents such as maintenance manuals, service instructions, and spare parts catalogues. No systematic method exists for extracting this information and mapping it onto a target knowledge graph schema. The integration of large language models (LLMs) for schema-guided information extraction represents an open research direction.

 

Objective

The objective of this thesis is to design, implement, and evaluate an LLM-based pipeline that automatically extracts operational disassembly knowledge from technical product documentation and populates the operational layer of an existing knowledge graph. Specifically:

  • Investigate current approaches for LLM-based information extraction in engineering and manufacturing contexts
  • Design and implement an extraction pipeline that uses the knowledge graph schema as a guiding constraint for the LLM
  • Identify extraction strategies 
  • Evaluate extraction quality against a manually populated baseline through entity-level metrics and graph-level query equivalence
Requirements
  • Interest in AI, natural language processing, and current trends in digitalization
  • Enthusiasm for sustainable production, disassembly, and remanufacturing
  • Self-motivation, independence, and reliability
  • Solid programming skills (Python required), experience with LLM APIs or NLP frameworks is a plus
  • Familiarity with knowledge graphs or graph databases (e.g., Neo4j) is desirable
  • Good English and German language skills

 

If interested, please contact:

M. Sc. German Bluvstein

Research group: Assembly Technology and Robotics

Phone: 089 289 15542

E-Mail: german.bluvstein@iwb.tum.de

Possible start
sofort
Contact
German Bluvstein
Room: 1303
Phone: +49 89 289 15542
german.bluvsteiniwb.tum.de
Announcement