Ontology-integrated tuning of large language model for intelligent maintenance

Peng Wang, John Karigiannis, Robert X. Gao

Research output: Contribution to journalArticlepeer-review

19 Scopus citations

Abstract

As new AI technologies such as Large Language Models (LLM) quickly evolve, the need for enhancing general-purpose LLMs with physical knowledge to better serve the manufacturing community has been increasingly recognized. This paper presents a method that tailors GPT-3.5 with domain-specific knowledge for intelligent aircraft maintenance. Specifically, aircraft ontology is investigated to curate maintenance logs with encoded component hierarchical structure to fine-tune GPT-3.5. Experimental results demonstrate the effectiveness of the developed method in accurately identifying defective components and providing consistent maintenance action recommendations, outperforming general-purpose GPT-3.5 and GPT-4.0. The method can be adapted to other domains in manufacturing and beyond.

Original languageEnglish
Pages (from-to)361-364
Number of pages4
JournalCIRP Annals
Volume73
Issue number1
DOIs
StatePublished - Jan 2024

Bibliographical note

Publisher Copyright:
© 2024 The Author(s)

Funding

The authors would like to thank Dr. Jianjing Zhang from Case Western Reserve University for his valuable contributions. Support from the National Science Foundation under awards CMMI- 2237242 , CNS- 2125460 , and EEC- 2133630 (Engineering Research Center on Hybrid and Autonomous Manufacturing – Moving from Evolution to Revolution) is sincerely appreciated.

FundersFunder number
National Science Foundation Arctic Social Science ProgramCMMI- 2237242, CNS- 2125460, EEC- 2133630

    Keywords

    • Large language models
    • Machine learning
    • Maintenance

    ASJC Scopus subject areas

    • Mechanical Engineering
    • Industrial and Manufacturing Engineering

    Fingerprint

    Dive into the research topics of 'Ontology-integrated tuning of large language model for intelligent maintenance'. Together they form a unique fingerprint.

    Cite this