Diagnostics of Engineering Systems using Large Language Models

Abstract

Intelligent fault diagnosis underpins the safe and reliable operation of modern engineering systems. While large language models (LLMs) have advanced intelligent diagnostic technologies for industrial scenarios, their real-world deployment remains constrained by domain semantic gaps, insufficient physical fault knowledge, and inherent hallucination issues. This Special Issue compiles 9 original studies covering multi-scale signal processing, multimodal fusion, physical-coupled fault modeling, knowledge-driven domain adaptation, semantic optimization, multimodal LLM construction, few-shot learning, and physics-verifiable structural health monitoring. Validated across diverse engineering systems, these works collectively connect foundational diagnostic methods with emerging LLM-enabled approaches. This editorial summarizes the key contributions of the collected papers, discusses current technical limitations, and outlines future research directions. It aims to promote the practical integration of LLMs within engineering system diagnostics.