Noise-robust machinery fault diagnosis based on self-attention mechanism in wavelet domain

Fault diagnosis of rolling bearings in complex environments is a difficult problem. First, the median filter can remove the noise in the vibration signals, however, it cannot adaptively adjust the filter weights according to the input signals. Second, the popular vision transformer (ViT) cannot extract local feature information under complex conditions and has a large number of parameters, which result in increased computational complexity. To solve these problems, a lightweight multi-feature fusion ViT bearing fault diagnosis method with strong local awareness in complex environments is proposed. Firstly, to learn the features and statistical distributions of the input signals, the gradient descent method is used to continuously and iteratively update the weights and filter the signals. Then, to better extract critical local fault information, a local sensing module is constructed using MWCNN. Finally, an improved lightweight multi-feature fusion ViT is constructed to perform global feature extraction and fault identification. The results show that the proposed method has better noise reduction effect and feature extraction ability, and can accurately identify the fault types under the complex environments.

Section: Experiments Results and Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A lightweight multi-feature fusion vision transformer bearing fault diagnosis method with strong local sensing ability in complex environments

Li,

Zhao

2024

“…Attention is an important technique that can help computer allocate resources more efficiently when dealing with tasks such as sequence learning, image recognition and speech understanding. It can operate adaptively between different modes by combining with layers that represent high-level abstractions [18]. SE block is a lightweight gating mechanism that can be used to enhance the representation of neural networks by modeling the relationship of the channels, thus plugging into the middle stage of the deep network.…”

Section: Sesfmentioning

confidence: 99%

Attention mechanism guided sparse filtering for mechanical intelligent fault diagnosis under variable speed condition

Han,

Wang,

Wan

et al. 2024

Variable speed is one of the common working conditions of mechanical equipment,which poses an important challenge to equipment fault diagnosis. The current solutions have the shortcomings of low computational efficiency and large diagnostic errors. The ability of attention mechanism to automatically extract useful features has begun to attract widespread attention in the field of mechanical intelligent fault diagnosis. Combining the advantages of attention mechanism and unsupervised learning, this paper proposes a squeeze-excitation attention guided sparse filtering (SESF) method for mechanical intelligent fault diagnosis method under variable speed. Firstly, the SE attention mechanism is embedded in SF algorithm to guide model training. Then, unsupervised feature extraction is carried out on the variable speed signal samples. The training results are adaptively screened and weighted to make the model pay more attention to the region with the most classify discrimination, so as to improve the feature extraction ability of the model to obtain useful information. Finally, two sets of gear and bearing tests under variable speed condition are adopted to testify the performance of the proposed method. The experimental results show that the SESF method can overcome the influence of variable speed to achieve accurate recognition of different mechanical faults and is superior to the other methods.

“…These methods are widely used, and after continuous improvement, they have achieved excellent results in bearing fault diagnosis under tasks such as noise and variable working conditions. For example, [21] proposed MANANR and designed two multi-scale noise reduction modules, which can adaptively eliminate noise in multi-scale convolution features; [22] proposed a method that combines RNN and GRU, its excellent anti-noise ability has been verified through experiments; [23] proposed Wavelet-SANet, which uses the self-attention mechanism and Transformer module to suppress noise in the wavelet domain and time-domain scattering impulse noise respectively. Li et al [24] improved the IICN and Inception network and finally realized the fault diagnosis of bearings with variable speed [25] Proposed MECNN, which uses Efficial channel attention to enhance features and combines CNN and Mode normalization to effectively solve the bearing variable load problem; [26] proposed a method based on CWT and T-ResNet, using transfer learning to effectively solve the problem of variable speed and load.…”

Section: Introductionmentioning

confidence: 99%

A small sample bearing fault diagnosis method based on novel Zernike moment feature attention convolutional neural network

Zhao,

2024

Bearings are one of the core components of rotating machine machinery. Monitoring their health status can ensure the safe and stable operation of rotating machine equipment. The limited nature of bearing fault samples makes it difficult to meet the demand for sufficient samples based on deep learning methods. Therefore, how to solve the problem of small- samples is the key to achieving intelligent fault diagnosis. In bearing failures based on vibration signals, the complex operating environment causes the vibration signals to inevitably mix with noise. The mixing of fault signature features and noise intensifies the strong spatial coupling of different types of fault features. In addition, diagnosing bearing failures under different loads is challenging because of the complex working conditions of bearings. Given the above problems, a small sample-bearing fault diagnosis method based on a high and low-frequency layered algorithm (HLFLA) and a novel Zernike moment feature attention convolutional neural network (ZMFA-CNN) is proposed. First, the proposed HLFLA converts one-dimensional time series signals into two-dimensional signals distributed rectangularly according to different frequency bands, and is used to simplify network feature screening, reduce the impact of noise, and retain adjacent signal constraint information. In addition, a new Zernike moment feature attention convolutional neural network is proposed to further extract multi-order moment features and attention weights, and can significantly improve the model generalization ability without increasing model parameters. At the same time, it is combined with FRN and TLU to further improve model performance. Finally, sufficient experiments verified that the algorithm proposed in this paper can solve the above problems and has excellent transfer generalization ability and noise robustness. In addition, the experimental results of applying the algorithm proposed in this article to gas turbine main bearing fault diagnosis prove the reliability of the algorithm proposed in this article.