Session: VIB-15-01: Machine Learning Applications in Vibrations and Dynamics

Paper Number: 88808

88808 - A Deep Long Short-Term Memory Network for Bearing Fault Diagnosis Under Time-Varying Conditions 

Rolling bearing, as one key functional element in the rotating machinery has been extensively utilized. Due to the harsh and long-time operating conditions, bearing is prone to failure. To maintain the bearing health and thus ensure the normal system function, fault diagnosis and prognosis of bearing is critically important. Machine learning-based fault diagnosis using vibration measurement recently has become a prevailing approach, which aims at identifying the fault through exploring the correlation between the measurement and respective fault. Nevertheless, such correlation will become very complex for the practical scenario where the system is operated under time-varying condition. To fulfill the reliable bearing fault diagnosis under time-varying condition, this study presents a tailored deep learning model, so called deep long short-term memory (DLSTM) network. By fully exploiting the strength of this model in characterizing the temporal dependence of time-series vibration measurement, the diagnosis performance can be improved. A publicly accessible dataset that was acquired under various time-varying conditions is utilized for illustration. A counterpart of DLSTM network, i.e., CNN is constructed to provide a baseline. Different testing cases are formulated, upon which the statistical validation procedures are executed to facilitate the thorough performance assessment. The results in case studies clearly indicate that the proposed DLSTM network outperforms the CNN in all testing cases in terms of classification accuracy, showing its effectiveness for bearing fault diagnosis under time-varying condition.

Presenting Author: Kai Zhou Michigan Technological University

Presenting Author Biography: Dr. Kai Zhou received the BS and MS degrees in Automotive Engineering from Chongqing University, Chongqing, China in 2007 and 2010, respectively. He obtained the PhD degree in Mechanical Engineering from University of Connecticut in 2015. After graduation, he worked as a research engineer at Bentley Systems and then joined University of Connecticut as a postdoctoral researcher. He joined Department of Mechanical Engineering and Engineering Mechanics in Michigan Technological University as a research assistant professor since 2020. His research interests include the physics-guided data-driven computational intelligence for structural health monitoring, uncertainty quantification and design optimization of structural dynamic system and manufacturing processes, and smart structures and sensing. He has 10 years+ academic and industriy experiences in relalated technology and research development.

Authors: