MMGCL: Multi-scale and multi-channel graph contrastive learning for flight anomaly detection

Nengjun Zhu, Yuqiang Ren, Yu Liu, Hang Yu, Xinzhi Wang, Xiangfeng Luo

Abstract: Highlights•The proposed MMGCL can detect flight anomalies and reveal the anomaly’s cause at sensor levels.•MMGCL can simultaneously model flight similarities and sensor correlations by multi-scale graph learning.•A multi-channel graph contrastive learning is proposed to differentiate normal and abnormal flight representations.•A novel two-step decision-making mechanism is proposed to determine whether the flight is abnormal.•Extensive experiments validate the superiority of MMGCL on interpretability and precision performance.

External IDs:dblp:journals/kbs/ZhuRLYWL25