ECE Students and Professor win IEEE Best Industry Paper Award
October 9, 2020
Electrical and Computer Engineering (ECE) Department graduates Jackson Cornelius and Blake Brockner, currently at CFD Research Corporation, along with Drs. Seong Hyeon Hong and Yi Wang from the University of South Carolina, Dr. Kapil Pant from CFD Research Corporation, and Dr. John Ball from Mississippi State University, won the best industry paper award at the 2020 IEEE International Conference on Prognostics and Health Management (ICPHM) for their paper entitled “Estimating and Leveraging Uncertainties in Deep Learning for Remaining Useful Life Prediction in Mechanical Systems”.
The ICPHM is the premier venue for scientists and engineers together and share cutting edge research related to health management and prognostics since 2009. Topics covered span from modeling and simulation to verification and validation; from analysis of uncertainty to integrated structural/frame health management; from Prognostics and Health Management (PHM) within distributed, networked, cloud computing, information, and communications systems to battery system health management.
The work performed by Dr. Ball and this team reported a novel approach for estimating epistemic and heteroscedastic aleatoric uncertainties in deep neural network models for remaining useful life prediction of mechanical systems. “This paper gives researchers deeper insights into data uncertainty and how the uncertainty affects remaining useful life predictions, which not only allows better analysis of engine failure (as shown in this paper), but also has wide applicability to failure assessment in general,” stated Dr. Ball.
“Dr. Ball has established himself as a leader in deep learning and its applications to myriad of fields including healthcare. The Department is proud of Dr. Ball’s latest achievement, and we all feel prouder of the ECE graduates being part of the award-winning team,” said Dr. Khan, the ECE Department Head.
For more information about Dr. Ball, please visit his webpage.