A Comprehensive Approach to Cataloging Missile Aerodynamic Performance Using Surrogate Modeling Techniques and Statistical Learning

摘要

A fundamental understanding of the aerodynamics of missile systems is foundational to missile system engineering. For many systems, the Mach number and angle of attack range can cover both attached and separated flows with and without interference from upstream artifacts through subsonic, transonic, supersonic and, occasionally hypersonic flight conditions. Wind Tunnel tests, flight tests, and computational fluid dynamics are all used in an effort to both fundamentally discover the nature of fluid flow around missile systems and to enable the reliable prediction of the integrated aerodynamic characteristics, which is critical to the dynamic simulation of missile system performance. In this paper, we develop a multivariable function approximation approach, using statistical learning techniques, such as, projection pursuit regression, neural networks and multivariate nonlinear regression, in support of rapid characterization of the aerodynamic characteristics of missile systems and show that it is a viable alternative to existing fast predictor methods.