At the Asia Pacific Conference of the Prognostics and Health Management Society (PHMAP) 2025, which took place in Singapore from December 8–11, 2025, Melanie B. Sigl presented the paper
„A Transfer Learning Framework for Remaining Useful Life Estimation“
The paper addresses the challenge of using robust deep learning models for remaining useful life estimation even with limited and incompletely annotated datasets. To this end, REAPER was developed – a framework that analyzes datasets and recommends suitable source models for transfer learning in order to avoid negative transfer. The goal is to make the selection of the initial model data-driven and transparent, rather than heuristic.
The discussions following the presentation emphasized the high relevance of this issue in the field of prognostics and health management (PHM). In particular, a generic and extensible framework that can be used across different datasets, architectures, and ranking models addresses the clearly perceptible need for practical solutions.