CCDC 2023
20-22 May

Distinguished Lectures

Prof. Aaron D. Ames, California Institute of Technology, USA

Researcher Long Cheng, Chinese Academy of Sciences, China

Prof. Na (Lina) Li, Harvard University, USA

Prof. Junzhi Yu, Peking University, China

Prof. Fumin Zhang, Georgia Institute of Technology, USA


Theoretical exploration and practice of industrial process fault diagnosis based on zero sample learning

Prof. Chunhui Zhao

Zhejiang University, China

Abstract

Fault diagnosis system is an important guarantee for the safe and reliable operation of industrial processes. Data-driven fault diagnosis modeling often depends on the collected sufficient historical fault data. However, in actual industrial processes, it is common that process failures have no historical samples and no labels. In this regard, we need to deal with a very challenging fault diagnosis task, that is, to consider diagnosing when there are no historical fault samples available for model training. We introduced the concept of zero-shot learning into the industrial field for the first time, and innovatively established a zero-shot-learning fault diagnosis method. By skillfully introducing a priori modeling knowledge with fault description as the carrier and adopting the attribute migration method, we overcame the concerned bottleneck problem that traditional fault diagnosis research cannot meet the sample size constraint. We theoretically analyzed and explained the effectiveness and feasibility of the zero-shot-learning diagnosis method based on fault description. In addition, the fault diagnosis effects in the case of zero samples are verified in the real megawatt thermal power process, and the results show that it is feasible to diagnose unseen target fault without samples. On this basis, the existing challenges, difficulties and future research directions are revealed.

Biography

Chunhui Zhao , Qiushi Distinguished Professor, Recipient of the National Outstanding Youth Fund. She received Ph.D. degree from Northeastern University, China, in 2009. From 2009 to 2012, she was a Postdoctoral Fellow with the Hong Kong University of Science and Technology and the University of California, Santa Barbara, Los Angeles, CA, USA. From 2012 to 2014, she was a distinguished researcher with Zhejiang University and since Dec. 2014, she has been a Professor with the College of Control Science and Engineering, Zhejiang University, Hangzhou, China.

Her research interests include statistical machine learning and data mining for industrial application. She has authored or coauthored more than 180 papers in peer-reviewed international journals. She has published 3 monographs and one national textbook. She authorized more than 60 invention patents. She is principal investigator of a Distinguished Young Scholar Program supported by the Natural Science Foundation of China. She has hosted more than 20 scientific research projects, including the NSFC funds, National key R&D project, provincial projects and corporate cooperation projects. She has received the Ministry of Education Natural Science Award and other provincial and ministerial awards. She also received more than ten academic awards, including the First Prize of Natural Science of Chinese Association of Automation, the First Young Women Scientists Award of Chinese Association of Automation, etc. She has served AE of three International Journals, including Journal of Process Control, Control Engineering Practice and Neurocomputing, and three domestic journals, including Control and Decision, etc.


Prof. Bin Zhou, Harbin Institute of Technology, China