Industrial Fault Diagnosis via Adversarial Collaboration Among Large Language Models

View at publisher Summarise with VIRAT
Authors: Ziming Li, Ding Wang, Chenxi Guo, Haoran Sun
Cited by: 0
References: 7

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper proposes an adversarial collaboration framework among multiple large language models (LLMs) for industrial fault diagnosis. While frontier models such as ChatGPT and Gemini demonstrate remarkable reasoning and generalization capabilities, their performance in processing localized industrial data-especially in non-English contexts such as Russian or Chinese-remains limited compared to domain-specific regional models. To address this challenge, several heterogeneous LLM agents are integrated to generate, challenge, and verify diagnostic hypotheses through adversarial interactions. A knowledge-grounded module based on retrieval-augmented generation (RAG) incorporates industrial manuals and maintenance logs to enhance contextual reasoning. Experimental results show that the proposed framework improves robustness, reduces false positives, and provides interpretable decision paths. The study demonstrates the potential of combining global and local LLMs in an adversarial collaborative manner for intelligent and multilingual fault analysis in Industry 4.0 systems.

Fields of Study & Subjects

Machine Fault Diagnosis Techniques Fault Detection and Control Systems Topic Modeling

Publication details

Pages
144–147
Publication date
3 February 2026
Language
EN