Daichao Wang

Reliable fault diagnosis of rotating machinery is essential for stable and safe operation of industrial systems. However, challenges such as significant data distribution discrepancies and intra-domain shifts caused by latent noise in the unlabeled target domain pose major barriers to the practical application of deep learning based fault diagnosis. These issues are particularly evident for samples near the decision boundary. Such samples often exhibit higher uncertainty and reduce the separability of the target feature space. To address these challenges, this paper proposes an uncertainty-aware domain adaptation framework for bearing fault diagnosis. First, we propose a perturbation-based module to measure sample uncertainty. It jointly uses perturbation sensitivity and historical information, enabling stable and reliable uncertainty estimation. Second, all domain alignment designs are adaptively guided by quantifying and exploiting sample uncertainty throughout training. For inter-domain alignment, we introduce an uncertainty-weighted discrepancy measure to improve the accuracy of cross-domain comparison. For the often-neglected intra-domain modeling, we further propose an uncertainty ranking based dual-matrix mechanism to improve pseudo-label reliability and preserve the geometric structure of the target domain. Experiments on a public dataset and an experimental bearing dataset demonstrate that the proposed method consistently outperforms state-of-the-art fault diagnosis methods in terms of diagnostic accuracy and reduce sample uncertainty.

This paper introduces a new method for fault diagnosis in rolling bearings, addressing performance drops caused by domain shifts under changing operational conditions. Unlike domain generalization, which relies on multiple source domains, this method focuses on single domain generalization to learn robust features from a single domain and generalize to new conditions. The proposed method, called multi-Gaussian attention-based single domain generalization (MGA-SDG), aims to enhance model generalization to unseen target domains. Multi-Gaussian attention projects multi-scale fault features into Gaussian feature spaces using both single modal and bimodal Gaussian functions, assigning attention weights based on feature alignment with these functions. This process ensures consistent and robust feature representations across domains. Furthermore, a covariance loss is employed to maintain distinct distributions for the weighted features, enhancing feature diversity. Experimental results on both public and proprietary rolling bearing datasets show that MGA-SDG achieves over 95.81% accuracy. This performance exceeds that of state-of-the-art methods, highlighting its potential for real-world industrial applications.

Bearing fault diagnosis is vital for saving invaluable time and cost since it is the most critical components in rotary machines. The feature fusion method has been a effective way to enhance the performance of fault diagnosis. However, how to extract and fuse the complementary fault features from multi-sensor data is still a problem to be solved. This study proposes a multi-sensor hybrid feature fusion network (MSHFFN) to fully mine the complementary fault information from multi-sensor signals for bearing fault diagnosis. In the network, the empirical features calculated by statistical methods and deep features are fused for fault classification, since more comprehensive representation of the bearing’s condition will be provided. Besides, A hybrid attention which consists of self attention and mutual attention is proposed innovatively for deep feature extraction. Paderborn bearing data set is used to verify the effectiveness of the MSHFFN. Results show that the diagnostic accuracy of MSHFFN can be up to 99.74%, which is significantly higher than other methods.

The bearing is the key component of rotating mechanical equipment, so the fault diagnosis of bearings is important to improve the reliability and safety of equipment operation. In recent years, feature fusion method has been extensively explored in the health monitoring and fault diagnosis of bearings. However, almost all the existing feature-fusion-based fault diagnosis methods extract features from different signals independently and concatenating them simply. It will lead to the failure of achieving the expected diagnostic accuracy because the complementary fault information is not fully mined and fused. This article proposes a novel bearing fault diagnosis approach based on mutual attention and bilinear model to address these issues. The features extracted from different input are interactive through mutual attention and are fused by the bilinear model, so the complementary fault features are effectively extracted and fine-grained fused. Experiments are conducted on the Paderborn bearing dataset to verify the effectiveness of the proposed method. Results show that the proposed method can effectively extract complementary fault features from different signals and deeply fuse them, and its diagnosis accuracy is up to 99.86%. Its performance is much better than that of simple concatenation and the feature fusion methods proposed in the reference.

Bearing is the key component of rotating machinery, so the fault diagnosis of bearing is important to improve the reliability of equipment operation. In recent years, feature fusion method has been extensively explored in fault diagnosis of bearings. However, the complementary fault features from multisensor data are difficult to be fully extracted, which will lead to the failure of achieving the expected diagnostic accuracy. This paper proposes a multitask network for bearing fault diagnosis. The multi-head attention is improved by 1D CNN to extract the deep features of multisensor data. The task of feature source discrimination allows the extracted features to contain complementary fault information as much as possible. Based on the complementary fault features, the accuracy of fault category classification task can be greatly improved. To verify the effectiveness of the proposed method, the experiments are conducted on Paderborn bearing data set. The results show that the accuracy of proposed method is greatly improved, which is much higher than the other methods.