기계공학부 구성원들의 많은 관심과 참여 부탁드립니다.

▣ 주   제: Uncertainty Reduction with Data in Prognostics and Health Management (PHM)

▣ 연   사: 김남호 교수

▣ 소   속: University of Florida

▣ 일   시: 2024. 12. 17.(화) 16:00

▣ 장   소: 제1공학관 A320호

▣ 초   청: 이종수 교수

▣ 초   록

 In the era of data science, some applications have too many data, while others have too small data. In practice, however, it is difficult to estimate how many data are good enough or what the effect of the lack of data would be. In addition, many students who learned physics and engineering are struggling to find their position in the flood of data science. Should we give up all the physics and engineering knowledge that were accumulated for more than thousands of years? This presentation focuses on how engineers should look at data and how to understand data from the viewpoint of engineers. Although there are numerous ways of looking at data, this presentation evaluates the contribution of data in the viewpoint of uncertainty reduction. If the source of uncertainty can be classified into aleatory and epistemic, we view data as a tool to reduce epistemic uncertainty. This unique view can tell us how many data are good enough and what we need to pay if we do not have enough data. Three related research topics will be presented: (1) prognostics under data deficiency, (2) data measurement scheduling, and (3) data-driven prognostics with low-fidelity physical information.