Author(s)

Chenfeng Jin, Jie Bai

Corresponding Author

Chenfeng Jin

Abstract

To enhance the accuracy of fault diagnosis in aircraft engines and improve the ability to capture critical information, this study proposes a model that integrates attention mechanisms with a Convolutional Neural Network and Bidirectional Gated Recurrent Unit (CNN-BiGRU). The Convolutional Neural Network (CNN) processes input data through multiple convolutional and pooling layers, effectively extracting spatial features. The BiGRU helps the model capture contextual dependencies, providing comprehensive dynamic analysis by processing both forward and backward data streams. The self-attention mechanism(Attention) enhances the focus on critical information in fault diagnosis through dynamic weight allocation. The integration of the CNN has significantly improved the model's feature extraction capabilities, likewise, the incorporation of the self-attention mechanism has strengthened its ability to capture essential information. Validated on NASA's C-MAPSS dataset, Compared with the most advanced models, the proposed framework achieves better performance. This model is not only efficient in training but also holds promising prospects for applications in fault prediction within the aviation field.

Keywords

Aero Engine Fault Diagnosis, Convolutional Neural Network, Bidirectional Gated Recurrent Unit, Self-Attention Mechanism