Event

ECE Colloquium - Dehong Lui, Mitubishi Electric Research Laboratories

Friday, November 6, 2026
3:30 p.m.
Jeong H. Kim Engineering Building, Room 1110
Darcy Long
301 405 3114
dlong123@umd.edu

Speaker: Dr. Dehong Lui, Research Scientist, Mitsubishi Electric Research Laboratories

Talk Title: Motor fault detection: from signal processing to machine learning

Abstract: Induction motors often operate under varying load and speed conditions to achieve high efficiency. However, these operating variations make motor fault detection challenging because changes in operating conditions can obscure or shift fault-related signatures.

In the first half of this talk, I will introduce a minimum-variance (MV)-based spectral analysis method to mitigate the effects of varying load conditions, together with a robust sparsity-driven method for estimating the unknown motor speed directly from measured current signals. The estimated speed is then used to compensate for speed-induced variations in the MV spectrum, enabling fault-related spectral signatures to be effectively extracted through an iterative process. The proposed approach is validated using experimental data collected from train traction motors under variable operating conditions.

In the second half of the talk, I will present a deep learning framework that combines deep generalized canonical correlation analysis (DGCCA) with attention-based multimodal fusion. The framework is evaluated using multimodal data collected from motors operating under a wide range of conditions. Experimental results demonstrate its capability to detect early-stage motor faults with subtle fault signatures and to estimate fault severity.

Bio: Dehong Liu received his B.S., M.S., and Ph.D. degrees all in electrical engineering from Tsinghua University, Beijing, China, in 1997, 1999, and 2002, respectively. From 2003 to 2010, he was with the Department of Electrical and Computer Engineering, Duke University, Durham, NC, USA, first as a Research Associate and then as a Senior Research Scientist. Since 2010, he has been with Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA, USA, where he is currently a Lead Research Scientist. His research has spanned a broad range of areas, including electric machines, power electronics, electromagnetic compatibility and interference (EMC/EMI), array signal processing, compressive sensing, image processing, and machine learning. His current research interests focus on electric machine fault diagnosis, signal processing, and machine learning.