Deep learning-based fault diagnosis methods typically require large amounts of fault data. To enable accurate gearbox fault diagnosis in small-sample scenarios, a novel diagnosis method based on a frequency band attention network is proposed.

First, a reconstruction-encoding layer is used to transform vibration signals into sub-band encoded signals that are more suitable for classification. Then, an intrinsic band attention layer is designed to effectively extract salient time-frequency features from the sub-band encoded signals. Finally, a multi-feature fusion module is used to integrate the extracted time-frequency features for fault recognition in small-sample conditions.

Experimental results on a gearbox fault simulation platform show that the proposed method achieves a fault diagnosis accuracy of 99.85% in small-sample conditions, surpassing existing benchmark models.

These findings can provide a valuable reference for gearbox fault diagnosis in small-sample conditions.

To address the challenge of low fault diagnosis accuracy in traditional neural networks with few labeled samples, a method based on contrastive learning and convolution transformer network is proposed.

First, raw monitoring data are transformed into similar sample pairs by data augmentation. These similar sample pairs are then mapped to a deep feature space by a feature extractor. A transformer network is utilized to design cross-prediction tasks for both local and global comparisons, facilitating the clustering of data with the same fault type by comparing the intrinsic similarity between the same batches of data. Finally, the downstream classification network is trained with few labeled samples to improve the diagnostic performance of the proposed model.

The effectiveness of the proposed method is validated using a self-built reducer test rig. The results show that accuracy of the proposed method reaches 98.38% with few labeled samples, showing significant advantages over existing methods.

The research results can provide the key technology for fault diagnosis of industrial equipment with few labeled samples, contributing to the advancement of intelligent manufacturing.

Deep learning has shown great potential in the field of rotating machinery fault diagnosis. Its excellent performance heavily relies on sufficient training samples. However, in practical engineering applications, acquiring sufficient training data is particularly difficult, resulting in poor generalization capability and low diagnostic accuracy. Therefore, few-sample fault diagnosis methods, which can effectively extract fault-related information from limited data, have gradually become a research focus in both academic and engineering circles.

In this paper, the latest achievements in few-sample fault diagnosis of rotating machinery are reviewed and summarized. This paper describes the definition and learning methods for few-sample fault diagnosis. Few-sample fault diagnosis methods aim to effectively develop fault diagnosis models with strong generalization capability under limited training data conditions. Currently, according to different technical principles, existing few-sample fault diagnosis methods can be classified into five categories: meta-learning, transfer learning, domain generalization, data augmentation, and self-supervised learning. Subsequently, this paper elaborates on the applications of these five methods in rotating machinery fault diagnosis. Meta-learning-based fault diagnosis methods improve the ability of models to rapidly learn and adapt to new tasks by acquiring common knowledge from multiple related tasks. The transfer learning-based fault diagnosis methods achieve knowledge migration from the source domain to the target domain using unsupervised domain adaptation techniques. The domain generalization-based fault diagnosis methods train models using single or multiple source domains and enable the model to learn features that are common across those domains. The data augmentation-based fault diagnosis methods expand the original dataset by generating models. The self-supervised learning-based fault diagnosis methods exploit the structural information of data to construct pseudo-labels.

The paper summarizes the core ideas, advantages, and limitations of these five methods. Meta-learning can improve the model's generalization capability but may require significant computational resources. Transfer learning can improve learning efficiency but is limited by domain similarity. Domain generalization can enhance the model performance in unknown domains but may suffer from overfitting issues. Data augmentation can increase dataset diversity but may generate inconsistent samples. Self-supervised learning can utilize unlabeled data but faces challenges such as complex task design and potential overfitting.

In the future, data governance, multimodal learning, federated learning, and mechanism-data hybrid-driven methods should be further explored in the field of few-sample fault diagnosis. It will overcome the limitations of existing methods and further improve the reliability of few-sample fault diagnosis.

Aiming at the problem of insufficient feature extraction in traditional neural networks under strong noise interference, a new global attention residual shrinkage network is proposed for accurate diagnosis of piston pump faults in complex environments.

First, data segmentation is performed on the original signals. Then, a new global feature extractor with an attention mechanism is established to extract fault-related features from the signals, while a threshold softening mechanism is introduced to minimize noise interference. Back propagation optimization is then performed on the network model to reduce loss and improve the model's diagnostic performance. Finally, the feature extraction results are input into the fault classifier for fault identification. The effectiveness of the proposed method is verified by using a piston pump fault simulation test bed.

The results show that, compared with other models, the established global attention residual shrinkage network model has higher diagnostic accuracy and stronger anti-interference ability.

The proposed method demonstrates accurate fault diagnosis in complex and harsh environments.

Aiming at problems including the fact that the fault diagnosis model of propulsion shafting systems under variable working conditions has poor generalization and cannot learn autonomously, and that the performance of the model is relatively fixed and cannot be updated online after it is deployed to the edge, this paper proposes a cloud-edge-end collaborative fault diagnosis method based on a deep residual shrinkage adaptive network.

First, the historical data of known operating conditions is collected and a deep residual shrinkage adaptive network model is built in the cloud through which reinforcement learning algorithms are introduced. These give the model the ability to update adaptively and learn data online under changing working conditions, thereby realizing online updating and adaptive performance enhancement. Model deployment and updating at the edge end are then realized by model slice distribution and edge slice aggregation, and real-time fault diagnosis is performed at the edge. Finally, the effectiveness of the proposed method is verified using a ship propulsion shaft system experimental bench.

The results show that the proposed method is able to realize the online updating of the model under variable operating conditions, and the updated model has higher fault diagnosis accuracy compared with a non-updated model.

The results of this study can provide useful references for the fault diagnosis of propulsion shaft systems under variable operating conditions.

Domain adaptive technology is widely used in the bearing fault diagnosis of variable operating conditions. However, most domain adaptive technology only focuses on the global domain distribution and ignores the subdomain distribution, and the domain-invariant feature quality is easily affected by noise, leading to a significant decrease in diagnostic accuracy under varying operation conditions. Therefore, a fault diagnosis method based on a self-attention subdomain adaptive adversarial network (SASAAN) is proposed.

First, a convolutional block attention module (CBAM) is utilized to extract the fault-related domain-invariant features in the vibration signals of the source and target domains. The adversarial network and subdomain adaptive module are then combined to reduce differences in the global and local domain edge distributions of different operating condition data, thereby improving the transferability of the data. The loss function is optimized by back propagation using the Adam optimizer to improve the diagnostic performance of the model, and the hyperparameter tuning of the model is also performed. Finally, the diagnostic results on the target domain test set are output by the failure classifier, and the Ottawa bearing data set is used to validate the effectiveness of the proposed method. ,

The results show that the fault diagnosis accuracy of the proposed method is higher than 96% under the condition of strong noise and varying operation conditions, which is obviously better than other methods.

The results of this study can provide valuable references for the fault diagnosis of rolling bearings under varying operation conditions.