A Safety‐Oriented Anomaly Detection and Self‐Healing O&M Framework for Aviation Edge Systems

OTA updates across aircraft, edge gateways, and ground systems can introduce faults that slow updates or affect service stability. This study builds an AIOps-based process that uses time-series data and log signals to detect OTA faults and to trigger automatic repair on aviation edge devices. Data were collected from 280 nodes over 90 days and included CPU load, memory use, network delay, and structured OTA logs. The detector reached an F1 score of 0.964 with a false-alarm rate of 1.1%, and most faults were identified within seconds. Automated actions resolved 73.6% of OTA issues and reduced manual tickets by 57.9%. These results show that combining simple sequence models, log features, and controlled rollback can shorten fault-location time and recovery time in cloud–edge–aircraft environments. The work provides a practical direction for improving OTA reliability, although wider testing across more aircraft types and update systems is still needed.