Abstract: Compared with manned vehicles,UAVs(Unmanned Aerial Vehicles) have many advantages,which make them widely used in military,civilian and scientific research fields.However,due to the lack of real-time decision-making ability,the UAV has high accident rate.Fault prediction is the core of UAV health management technology.Before building a fault prediction model,an important step is to identify the pattern of sampled data so as to add accurate labels to training data for modeling,which is also a part of improving flight portrait.Based on the UAV flight data accumulated in a big data platform of an UAV production company in Shenyang,this paper proposed a semi-supervised clustering technique to automatically identify the normal points of the flight process,the fault points(including the crashing points) and the points after crashing.At the same time,the management and statistics are strengthened,and the efficiency and accuracy of adding a precise label to the historical flights data are greatly improved.Real flight data or flight test data were used to verify the results.The results of manual verification show that the recognition rate of fault points can reach over 80%.

Key words: Fault prediction, Pattern recognition, Semi-supervised clustering, Unmanned aerial vehicles