A Multidimensional Feature Fusion Network for Fault Diagnosis of High-Speed Train Bogie Under Imbalanced Data

Ensuring the normal operation of high-speed train (HST) bogie is of great significance for the safe operation and maintenance of HSTs. With the development of deep learning, intelligent fault diagnosis of bogie has become a research hotspot in this field. However, data imbalance limits the diagnostic accuracy of the intelligent model. Therefore, this paper proposes a multidimensional feature fusion network. Firstly, the fusion of spatiotemporal information is achieved by jointing 1-D CNN and 2-D CNN. Secondly, coordinate attention mechanism is introduced to enhance the sensitivity of the network to effective features. Finally, the performance of the proposed network for bogie imbalanced fault diagnosis is verified using the dataset from SIMPACK simulation software. On the whole, the proposed network achieves an average accuracy of 93.39% on the unbalanced bogie dataset, which is superior to existing bogie fault diagnosis methods.