Effective condition monitoring and fault diagnosis of bearings can not only maximize the life of rolling bearings and prevent unexpected shutdowns caused by equipment failures but also eliminate unnecessary costs and waste caused by excessive maintenance. However, the existing deep-learning-based bearing fault diagnosis models have the following defects. First of all, these models have a large demand for fault data. Second, the previous models only consider that single-scale features are generally less effective in diagnosing bearing faults. Therefore, we designed a bearing fault data collection platform based on the Industrial Internet of Things, which is used to collect bearing status data from sensors in real time and feed it back into the diagnostic model. On the basis of this platform, we propose a bearing fault diagnosis model based on deep generative models with multiscale features (DGMMFs) to solve the above problems. The DGMMF model is a multiclassification model, which can directly output the abnormal type of the bearing. Specifically, the DGMMF model uses 4 different variational autoencoder models to augment the bearing data and integrates features of different scales. Compared with single-scale features, these multiscale features contain more information and can perform better. Finally, we conducted a large number of related experiments on the real bearing fault datasets and verified the effectiveness of the DGMMF model using multiple evaluation metrics. The DGMMF model has achieved the highest value under all metrics, among which the value of precision is 0.926, the value of recall is 0.924, the value of accuracy is 0.926, and the value of F1 score is 0.925.
Copyright © 2023 He-xuan Hu et al.