Effective Convolutional Transformer for Highly Accurate Planetary Gearbox Fault Diagnosis

Abstract

To extract the global temporal correlations and local features together to enhance the accuracy for fault diagnosis, this paper proposes an effective convolutional Transformer (ECT), which can learn the global temporal correlations using Transformer and local features with convolution at the same time. The proposed method designs a multi-stage hierarchical structure of Transformer, which utilizes convolutional tokenization to distill dominating sequence features from raw vibration signals while increasing the dimension of token embedding across stages at the same time as that in CNNs. The spatial-reduction attention (SRA) and the linear dimension reduction projections are introduced respectively to Transformer at different stages to reduce the resource consumption of the model. Finally, the proposed method utilizes a sequence pooling strategy on the output of Transformer to eliminate the requirement of the class token and make the model accurate for classification. The specially designed structure makes the model flexible and effective for planetary gearbox fault diagnosis. Experiments performed on planetary gearbox fault simulators indicate that the ECT method has significant effectiveness and high accuracy compared with the state-of-the-art methods for planetary gearbox fault diagnosis.

Playing Flappy Bird Based on Motion Recognition Using a Transformer Model and LIDAR Sensor

“…The dimension of this transformed output equals the dimension of the original input vector. Typically, the first layer of the block has 4 times more neurons than the last layer of the block [ 33 ].…”

Conditional Enhanced Variational Autoencoder-Heterogeneous Graph Attention Neural Network: A Novel Fault Diagnosis Method for Electric Rudders Based on Heterogeneous Information

“…The performance of the ER is crucial since it directly influences the overall functionality of the controlled system, particularly for hypersonic aircraft, and impacts control accuracy and hit rates [ 6 ]. For most mechanical devices, such as motor bearings, self-priming centrifugal pumps, and axial piston hydraulic pumps [ 22 , 23 , 24 , 25 , 26 ], FD typically entails the analysis of sensor data collected during actual operations. However, given the stringent safety, controllability, and repeatability requirements, along with the limitations in the quantity and quality of field tests for hypersonic aircraft, extensive research has been conducted on ground-based semi-physical simulation technology.…”

A review on deep learning in planetary gearbox health state recognition: methods, applications, and dataset publication

“…To exploit the merits of both CNN and transformer, Han et al [124] proposed a Convformer-NSE to extract the local and global information from raw vibration signals, in which two convolutional layers were added to process input data and a novel Senet (NSE) [125]was integrated with transformer to make full use of learning of channel and spatial adaptivity. Sun et al [126] designed a multi-stage hierarchical structure via convolutional tokenization for transformer to learn both local and global information from raw signals, and meantime, spatial-reduction attention and linear dimension reduction projections were introduced to transformer to reduce the resource consumption. Li et al [118] proposed a variational attentionbased transformer network (VATN) as shown in figure 15, in which the basic transformer model was improved by a sparse constrain to embed prior knowledge into the model.…”