A collaborative approach based on large language model and knowledge graphs for information integration towards smart manufacturing

Li, Ruihao, Chen, Chong, Liu, Ying ORCID: https://orcid.org/0000-0001-9319-5940, Wang, Tao, Shao, Haidong and Cheng, Lianglun 2026. A collaborative approach based on large language model and knowledge graphs for information integration towards smart manufacturing. Engineering Applications of Artificial Intelligence 176 (Part 2) , 114785. 10.1016/j.engappai.2026.114785

PDF - Accepted Post-Print Version Available under License Creative Commons Attribution. Download (3MB)

Abstract

In the era of smart manufacturing, integrating vast amounts of information has become an essential task. Knowledge Graph (KG) is a key technology for improving information integration, which can greatly improve the performance of question-answering for Large Language Models (LLMs). However, the existing approach mainly adopts KG as the plug-in database for Retrieval-Augmented Generation (RAG), which cannot achieve accurate answering due to the imperfections of KG. In order to address this challenge, a collaborative LLM-KG framework is proposed to iteratively update the KG, which can provide fine-grained knowledge for RAG. The methodology firstly constructs a foundational ontology, and adopts LLM for knowledge triples extraction to establish an initial KG based on multi-source data. Then, competency questions (CQs) are designed for the evaluation and optimization of the initial KG. After ontology optimization, a fine-grained KG is obtained to facilitate a robust question-answering mechanism through RAG. The proposed iterative approach can effectively refine the system's decision-support capabilities. An experimental study based on the real-world shipbuilding process data is implemented. The experimental results demonstrate that the answering accuracy can be improved from 86.18% to 93.09% with the enhancement of the proposed approach.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Schools > Engineering
Additional Information:	RRS policy applied
Publisher:	Elsevier
ISSN:	0952-1976
Date of First Compliant Deposit:	15 April 2026
Date of Acceptance:	6 April 2026
Last Modified:	15 Apr 2026 09:30
URI:	https://orca.cardiff.ac.uk/id/eprint/186391

Actions (repository staff only)

Edit Item