19th Asia-Pacific Vibration Conference (APVC2021)

325 / 2021-06-21 17:14:27

Bearing Fault Diagnosis Under Multiple Loads Based on Deep-Stacked CNN

Deep Learning; CNN; Bearing; Fault Diagnosis

With the continuous development of industries, rotating machinery has played an important role. However, it suffers from many disturbances and injuries during its operation. The relatively serious damage to rotating machinery is self-vibration, which occurs when gears and bearings fail. Gear vibration signals are often used for fault diagnosis to reduce loss. Given that rotating machinery often operates in complex production environments, the collected signals also have a high degree of complexity. Based on the convolutional neural network model, we broaden the feature dimension by stacking features of different dimensions and then extract the maximum value of the original data before finally gathering the features extracted by the two. To verify the generalization ability of the proposed model, this study uses the Western Reserve University dataset. Experiments are conducted with the drive-end fault signals loaded at 0, 1, 2, and 3 HP. Results are visually analyzed through T-SNE and confusion matrix.