Department News

Professor Byeng Dong Youn's research team Chan Hee Park and Yunhan Kim has won the PHM Asia Pacific 2021 Best Student Award

Author
yeunsookim
Date
2021-10-25
Views
508


 

Seoul National University's Department of Mechanical Engineering, Laboratory for System Health & Risk Management (advisor Professor Byeng Dong Youn), Ph.D. program students Chan Hee Park and Yunhan Kim received Best Student Award (Gold medal, Bronze medal respectively) at Asia Pacific Conference of the Prognostics and Health Management Society 2021 (PHM Asia Pacific 2021). The conference was held at the Ramada Plaza Jeju Hotel September 8 (Wed)-11 (Sat), 2021.

The thesis title of Chan Hee Park's doctoral program is “A Novel Health Feature for Fault Diagnosis of a Servo Motor Under Non-stationary Conditions”, and a new soundness image was proposed for the diagnosis of servo motor failures. Existing time-frequency analysis and imaging techniques of time-series signals suppress the operation-related components of the phase current signal in order to overcome the limitations in terms of fault diagnosis, such as lack of physical meaning or loss of fault-related components during the conversion process. a method to highlight and define fault-related components as images is proposed. In addition to evaluating the health of the proposed image itself, the convolutional neural network failure diagnosis model learned from it was recognized for its originality in that it enables physics-converged deep learning-based failure diagnosis.

The thesis title of Yunhan Kim's doctoral program is “A Novel Multiscale Convolutional Neural Network for Industrial Gearbox Fault Diagnostics” and proposed a new deep learning model for fault diagnosis of industrial gearboxes. Focusing the existing physics-based time-frequency analysis technology, by replacing the fixed form of the basis function with a learnable filter, convolution with effective time-frequency analysis for fault diagnosis A neural network model was proposed. The proposed method is a deep learning technology with a physics-based time-frequency analysis, and the originality and excellence of the thesis have been recognized in that it is possible to diagnose the failure of an industrial gearbox with a complex operating environment without relying expert knowledge.