Deep Learning for Gust Detection in Unsteady Aerodynamics

Recent successes of deep learning in various fields aroused the interest to integrate such algorithms into aerodynamics. Here, we present several deep learning structures to identify the unknown characteristics of incident gusts and rapid changes of angle of attack in the measured surface pressures on a flat plate. The flow field is simulated with a two-dimensional inviscid vortex model, in which the separated flow is modeled by the leading edge suction parameter (LESP) criterion. When the instantaneous value of LESP exceeds a varying critical value, the model releases vorticity proportional to the excess to simulate the effect of the gusts. We present several deep learning algorithms that use the surface pressure history to determine the characteristics of disturbances. The first algorithm aims to detect the LESP history at a constant angle of attack. The second algorithm aims to detect the angle of attack history during a pitch-up maneuver with constant critical LESP. The third and fourth algorithms concern mixtures of these two scenarios. The four algorithms achieve satisfactory results in their own tasks, showing the possibility of employing deep learning on a broader scale for gust detection in aerodynamics.