讲座人：
Dr. Zhenghua Chen is currently a Scientist, PI, PhD Supervisor and Lab Head at Institute for Infocomm Research, Agency for Science, Technology and Research (A*STAR), Singapore. He received the B.Eng. degree from University of Electronic Science and Technology of China and Ph.D. degree from Nanyang Technological University (NTU), Singapore.His research interests include time series data analytics, transfer learning, model compression and related applications (e.g., smart manufacturing). He works as PI and Co-PI for a number of research grants, including A*STAR CDA, AME Programmatic, NRF YIRG, etc. (Total amount > 10M SGD). He has published more than 90 top journals/conferences, such as IEEE TNNLS, Tcyber, TII, AAAI, IJCAI, ICCV, NeurIPS, etc., with a Google Scholar citation of 4800+. He has been listed as World’s Top 2% Scientists by researchers from Stanford University. Besides, he has won several competitive awards, such as A*STAR Career Development Award, First Place Winner for UG2+ Prize Challenge at CVPR 2021, First Runner-Up Award for Grand Challenge at IEEE VCIP 2020, Finalist Academic Paper Award of IEEE ICPHM 2020, etc. He currently serves as the Associate Editor for Neurocomputing and IEEE Transactions on Instrumentation and Measurement. He is also the IEEE Senior member and the Vice Chair of IEEE Sensors Council Chapter, Singapore.
讲座简介：
Machine health monitoring (MHM) is of great importance mechanical systems, such as vehicles, trains, airplanes, etc. It is able to reduce the maintenance cost and improve the reliability of the systems. A key task in MHM is the prediction of machine remaining useful life (RUL). However, due to the high complexity of modern mechanical systems, accurate prediction of machine RUL is very challenging. Artificial intelligence which is capable of modelling highly non-linear and complex systems can be a good candidate. In this presentation, we attempt to address three fundamental questions in RUL prediction via proposing efficient and adaptive AI models: 1) how to achieve accurate RUL prediction with advanced machine learning methods? 2) how to achieve RUL prediction on edge devices which have limited computational resources? 3) How to deal with cross-domain RUL prediction? To solve these challenging, but practical issues, we proposed an attention based LSTM method for accurate RUL predication, an efficient KDnet-RUL framework for compressing deep neural networks for RUL prediction on edge devices, and an adaptive cross-domain RUL perdition algorithm. Extensive experiments have been conducted to verify the effectiveness of the proposed methods.
讲座时间：2022/10/13 16:00-17:30 #
腾讯会议：502-770-531
腾讯会议链接: https://meeting.tencent.com/dm/8AEc8SrOgWb1
