ROLLING BEARING DIAGNOSIS WITH 1D TEMPORAL CONVOLUTIONAL NETWORK

Deep Learning (DL) has contributed a lot in the field of industrial maintenance, in particular predictive maintenance by detecting potential failures and breakdowns before their appearance. Unfortunately, the DL has some limitations like the need for a raw data preprocessing to produce an effective prediction model and also there is the problem of the model fragility in the face of changes in operating conditions. In this paper, we propose a fast and accurate condition monitoring and early fault detection system using a type of CNN architecture, 1D Temporal Convolutional Network (TCN), which has the capacity to fuse the feature extraction and classification phases of the bearing fault detection into one single learning body. The proposed approach is directly applicable to raw vibration data and therefore removes the need for a separate feature extraction step; resulting in more efficient system. The prediction model provides a fast and highly accurate anomaly detection and condition monitoring system. It showed a high precision and a strong degree of robustness against changes of functioning conditions.