TY - GEN
T1 - Data-Driven Fault Diagnosis Method for UAV Sensors Based on Spatiotemporal Cross Attention
AU - Li, Zhonghan
AU - Zhang, Yongbo
AU - Shi, Yutong
AU - Guo, Jianchao
AU - Zhu, Shihao
AU - Wang, Ling
N1 - Publisher Copyright:
©2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Unmanned aerial vehicles (UAV) sensor fault diagnosis faces the challenges of imbalanced class and data coupling. This paper proposes a novel data-driven method, Parallel CNN-LSTM with Cross Attention (PCL-CA), which enhances the model's feature extraction capability by introducing spatiotemporal cross-attention. A parallel architecture is designed, and enhanced data is generated using multi-rate EKF, ensuring the model achieves excellent fault diagnosis performance even in the presence of imbalanced class. Experimental results show that the proposed PCL-CA method exhibits the most robust performance across four performance metrics, achieving a precision of 92.8% and an F1 score of 93.1, significantly outperforming the CNN, LSTM, BiLSTM and CNN-TransNet baseline models.
AB - Unmanned aerial vehicles (UAV) sensor fault diagnosis faces the challenges of imbalanced class and data coupling. This paper proposes a novel data-driven method, Parallel CNN-LSTM with Cross Attention (PCL-CA), which enhances the model's feature extraction capability by introducing spatiotemporal cross-attention. A parallel architecture is designed, and enhanced data is generated using multi-rate EKF, ensuring the model achieves excellent fault diagnosis performance even in the presence of imbalanced class. Experimental results show that the proposed PCL-CA method exhibits the most robust performance across four performance metrics, achieving a precision of 92.8% and an F1 score of 93.1, significantly outperforming the CNN, LSTM, BiLSTM and CNN-TransNet baseline models.
KW - UAV sensors
KW - data driven
KW - fault diagnosis
KW - imbalanced class
KW - spatiotemporal cross attention
UR - https://www.scopus.com/pages/publications/105030473118
U2 - 10.1109/ICMAE66341.2025.11277186
DO - 10.1109/ICMAE66341.2025.11277186
M3 - 会议稿件
AN - SCOPUS:105030473118
T3 - 2025 16th International Conference on Mechanical and Aerospace Engineering, ICMAE 2025
SP - 242
EP - 247
BT - 2025 16th International Conference on Mechanical and Aerospace Engineering, ICMAE 2025
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 16th International Conference on Mechanical and Aerospace Engineering, ICMAE 2025
Y2 - 15 July 2025 through 18 July 2025
ER -