Bearing Fault Diagnosis Based on an Enhanced Image Representation Method of Vibration Signal and Conditional Super Token Transformer

Abstract

Multipoint Optimal Minimum Entropy Deconvolution Adjusted (MOMEDA) is an advanced deconvolution method, which can effectively inhibit the interference of background noise and distinguish the fault period by calculating the multipoint kurtosis values. However, multipoint kurtosis (MKurt) could lead to misjudgment since it is sensitive to spurious noise spikes. Considering that L-kurtosis has good robustness with noise, this paper proposes a multipoint envelope L-kurtosis (MELkurt) method for establishing the temporal features. Then, an enhanced image representation method of vibration signals is proposed by employing the Gramian Angular Difference Field (GADF) method to convert the MELkurt series into images. Furthermore, to effectively learn and extract the features of GADF images, this paper develops a deep learning method named Conditional Super Token Transformer (CSTT) by incorporating the Super Token Transformer block, Super Token Mixer module, and Conditional Positional Encoding mechanism into Vision Transformer appropriately. Transfer learning is introduced to enhance the diagnostic accuracy and generalization capability of the designed CSTT. Consequently, a novel bearing fault diagnosis framework is established based on the presented enhanced image representation and CSTT. The proposed method is compared with Vision Transformer and some CNN-based models to verify the recognition effect by two experimental datasets. The results show that MELkurt significantly improves the fault feature enhancement ability with superior noise robustness to kurtosis, and the proposed CSTT achieves the highest diagnostic accuracy and stability.

Bearing fault diagnosis method based on angular domain resampling, relative position matrix and transfer learning

“…Hou et al [45] proposed a fault diagnosis method based on the Gramian angular difference field and the transfer learning in the ResNet34 model, which was developed by encoding a one-dimensional vibration signal into a twodimensional image and utilizing a residual network for automated feature extraction and classification. Li et al [46] proposed a bearing fault diagnosis framework, which utilizes a multipoint optimal minimum entropy deconvolutional adjustment method and a conditional Super Token Transformer for feature extraction and classification, which has higher noise robustness and diagnostic accuracy than traditional methods. Zhao et al [47] proposed a bearing fault diagnosis method based on the Markov transfer field and CNN, which encodes the vibration signals and generates feature maps, and then utilizes the CNN for feature extraction and classification.…”

Multimodal Transformer for Bearing Fault Diagnosis: A New Method Based on Frequency-Time Feature Decomposition

“…For example, Li et al proposed the multipoint envelope L-kurtosis (MELkurt) method to extract time-domain features of bearing signals and transformed them into images using the Gramian Angular Difference Field (GADF) method. By employing the Conditional Super Token Transformer (CSTT), they achieved higher diagnostic accuracy and reliability [22]. Choudhary et al extracted bearing signal features using a noninvasive thermal image-based method and employed CNN based on the LetNet-5 structure for bearing fault classification [23].…”