Airspeed Anomaly Detection of UAV Based on Flight Mode Adaptive with Noise Margin

Unmanned aerial vehicles (UAVs) has been widely used. In order to ensure the completion of the mission, prognostic and health management (PHM) of UAV has been focused on. Airspeed anomaly detection is an important part of UAV PHM. However, most of the existing methods only use single model for different flight modes or phases. This will reduce the accuracy of the model. Therefore, in this paper, a set of airspeed anomaly detection models based on flight mode adaption with noise margin which will further improve the accuracy of precision are raised. This method estimate the airspeed with different Long Short-Term Memory (LSTM) Recurrent Neural Network based the flight mode. In order to detect the faults, PauTa criterion is applied. To verify the effectiveness of the method, the models are deployed on real flight data. The result of the experiment proves that the method is feasible.