(사)국제융복합연구원

키워드

초록

This study proposes a multi-label learning-based Prognostics and Health Management (PHM) system for enhancing reliability in military vehicles and unmanned robots. These systems, operated in harsh operational environments, frequently experience compound faults across various components simultaneously. Traditional single-label diagnostic models classify only one fault type at a time, thus risking overlooking some faults in compound failure situations. In contrast, multi-label learning assigns multiple fault labels to a single input, enabling comprehensive diagnosis of overlapping fault conditions. This research presents an integrated PHM architecture encompassing sensor data collection and preprocessing, multi-label classification-based fault diagnosis, state prediction, and maintenance decision support. The system’s effectiveness and limitations are analyzed through simulation-based experimental design. The multi-label learning approach is expected to detect compound faults without omission, thereby reducing post-failure response costs and unexpected mission downtime while improving operational reliability of military vehicles and unmanned robots.

참고문헌 (0)