AI(Artificial Intelligence) models specialized for a particular aircraft may not be suitable for other aircraft data because they are optimized for that aircraft's data distribution and pattern. In this paper, we attempted to increase the generalization ability of AI models for aircraft by proposing a new data preprocessing method based on the aircraft's equation of state using domain knowledge about aircraft. The simulation results of the AI model applying the proposed preprocessing method show that it can be universally applied to small UAVs by accurately determining whether to enter an abnormal flight state even with the data of the aircraft type that was not used for learning.