Special paper sessions are most welcome. Please use the template for special paper session to make a proposal. This template is downloadable from the “Downloads” page at this conference website. If you have any questions, please feel free to contact the conference organization committee by email: email@example.com.
Special Paper Session
Deep Learning for PHM: Opportunities and Challenges
Outline of the Special Paper Session
Deep Learning has benefited from several break-throughs in recent years, with advances in optimization and algorithmic techniques which, combined with progress in computing power and the increased availability of (labelled) datasets, have made their implementation practical. As a result, impressive successes have been achieved in various areas, such as computer vision, natural language processing, autonomous vehicles, etc.
Yet, despite the progress achieved in other domains, deep learning applications in PHM are still rather rare. There are several parallels in challenges faced by PHM applications and those which the computer vision community is currently working on to overcome: this includes lack or a limited availability of labels, domain specificity, and limited representativeness of the available training datasets. On the other hand, one of the driving forces of deep learning in computer vision was the introduction of benchmark datasets, such as ImageNet, which have significantly driven the developments in the domain. Something similar is still lacking in the PHM community, even though such datasets as C-MAPSS are heavily used by the researchers worldwide.
The goal of this session is to discuss the opportunities and hurdles of deep learning applications in the PHM domain, to look at new developments and how these developments are overcoming the concerns and limitations faced by PHM applications, and to see which new opportunities they open. A further goal is to discuss what PHM community can learn from other domains and which problems are specific to the PHM domain and still need to be overcome. A specific focus will be given to making these algorithms and models ready for industrial applications and how to drive the innovation for new developments in this field.
Topics of the Special Paper Session
Traditional machine learning approaches in PHM in comparison to deep learning approaches
Organizers of the Special Paper Session
Pierre Dersin is RAM (Reliability-Availability-Maintainability) Director and PHM (Prognostics & Health Management) Director of ALSTOM Digital Mobility. With ALSTOM Transport since 1990, with expertise and experience in railway RAM and maintenance, he founded the “RAM Center of Excellence” and initiated the PHM activity at ALSTOM. He holds a Ph.D. in Electrical Engineering and a Master’s degree in Operations Research, both from MIT. He has authored numerous publications (among other in IEEE Transactions, ESREL, RAMS, French Lambda-Mu symposia, IEEE-PHM Conferences). He served on the IEEE Reliability Society AdCom from 2012 to 2017 and is a member of the IEEE Future Directions Committee. He contributed four chapters in the “Handbook of RAMS in Railways: Theory & Practice” (CRC Press, 2018). His research interests include hybrid PHM methods; the application of machine learning and in particular deep learning; links between PHM and reliability engineering; reliability of systems of systems; and dynamic optimization methods applied to maintenance; with strong emphasis on railway applications.
Olga Fink is SNSF (Swiss National Science Foundation) professor for intelligent maintenance systems at ETH Zürich. Before joining ETH faculty, she was heading the research group “Smart Maintenance” at the Zürich University of Applied Sciences (ZHAW). Olga received her Ph.D. degree in civil engineering from ETH Zurich, and Diploma degree in industrial engineering from Hamburg University of Technology. She has gained valuable industrial experience as reliability engineer for railway rolling stock and as reliability and maintenance expert for railway systems. Olga’s research focuses on Data‐Driven Condition‐Based and Predictive Maintenance, Deep Learning and Decision Support Algorithms for Fault Detection, Diagnostics and Prognostics of Complex Industrial Assets.