2022 article

Store Separation Trajectory Clusters from Machine Learning

JOURNAL OF AIRCRAFT, Vol. 59, pp. 117–125.

By: W. Gothard n & K. Granlund n 

co-author countries: United States of America πŸ‡ΊπŸ‡Έ
author keywords: Gaussian Mixture Models; Shear Layers; Flow Conditions; Aerodynamic Properties; Karman Vortex Street; Wind Tunnel Walls; Cluster Analysis; Fluid Structure Interaction; Probability Distribution; Dynamic Time Warping
Source: Web Of Science
Added: March 21, 2022

Store separation of a generic, thin-finned, missile through a continuously oscillating shear layer into subsonic flow was conducted experimentally through 100 drop tests to identify potential groups of trajectories and statistical phenomena. Change in store pitch was observed using a high-speed camera. Trajectories were grouped using machine learning with a -means clustering, followed by a Gaussian mixture model clustering approach. The -means clustering revealed two primary groups and one outlier group. The statistical strength of the primary groups was confirmed with the Gaussian mixture model, which places 89% of trajectories into one of two groups. The existence of two primary groups is strong evidence of a bifurcation.