A CNN-based Airborne Weather Risk Warning Method

In this study, A CNN-based airborne weather risk warning method is proposed. This method can autonomously predict weather risk according to airborne weather radar data during flight. We use CNN to construct a weather risk prediction model, and convert airborne weather radar data into RGB images for risk identification. The model can provide weather risk warning when the pilots are distracted or have a heavy workload, and it has the advantages of high reliability, high efficiency, little computing power requirement, and can be stably run on the airborne deep learning inference platform. We perform the experiment using a Airbus A321 airplane model in a flight simulator, and the experiment results show that the weather risk prediction model can quickly and accurately identify weather risks, with an accuracy of 97.5%. We analyze the cases of misidentification and find that most of the cases belong to the middle state between risk and no risk.