| Title |
Low-latency Bearing Fault Diagnosis based on Convolutional LSTM Model |
| Authors |
김지호(Ji-Ho Kim) ; 신지훈(Ji-Hoon Shin) ; 김태환(Tae-Hwan Kim) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.1.124 |
| Keywords |
Fault diagnosis; Neural networks; Embedded systems; Low-latency monitoring systems |
| Abstract |
This paper presents a novel approach of bearing fault diagnosis based on convolutional LSTM (ConvLSTM) model to reduce the latency. Time-series sensor data are segmented into short vectors that are fed sequentially into a ConvLSTM model to find the spatio-temporal features effectively. The model is devised to a many-to-many structure by which the failure can be diagnosed as soon as several consecutive prediction results correspond to the failure condition. The proposed approach reduces the latency by up to 99.3% and 50.7% compared to the 2D-CNN-based and 1D-CNN-based approaches, respectively, without any degradation of the diagnosis accuracy. |