Yiliu Liu

This presentation, based on research published in Engineering (2025), introduces LLM4M, a domain-specific Large Language Model designed to optimize maintenance decision-making for wind farms. As global wind capacity is projected to reach 185.0 GW by 2026, the high cost of offshore operation and maintenance (O&M) has become a critical barrier to competitiveness. In this study, the LLM4M model based on labeled-data-supervised fine-tuning of the Qwen2.5-7B is developed to address the limitations of computationally intensive conventional methods. It utilizes low-rank adaptation to internalize complex spatial and logical constraints and proposes a five-step process involving information extraction, slot definition, and prompt construction to inject exact failure types and geographic coordinates into the model. Experimental results demonstrate that LLM4M significantly outperforms general-purpose models and traditional optimization approaches. LLM4M achieved a 100% execution rate for valid, implementable plans, compared to just 83.40% for non-fine-tuned baseline LLM. Cost Research also indicates that models trained on complex scenarios (≥ 3 teams) intuitively abstract universal routing rules, allowing them to solve simpler configurations effortlessly. 

Biography:

Yiliu Liu is a professor at the Department of Mechanical and Industrial Engineering, Norwegian University of Science and Technology (NTNU). His main research interests include system reliability and resilience, safety-critical system and safety barrier analysis, and risk-informed decision-making etc. The applications of his research are mainly within the energy and transportation sectors. He has published more than 200 papers on international journals and conferences. Currently, Prof. Liu is a member of Norwegian Academy of Technical Science (NTVA), the co-chair of Sweden-Norway joint section of IEEE Reliability Society, the editor-in-chief for International Journal of Reliability and Safety, the leader of NTNU NATECH (natural-hazard triggered technical accident) thematic research group, and the director of International RAMS (reliability, availability, maintainability, and safety) master program at NTNU.