Proper Orthogonal Decomposition and Response Surface Method for TPS/RLV Structural Design and Optimization: X-34 Case Study

Abstract

To improve the prediction of lee-side aerodynamics/aerothermodynamics for TPS/RLV structural design and optimization, a POD/RSM-based s/hypersonic aerodynamic module is developed from the high-level CFD solutions for the selected X-34 configuration. After the creation of an extensive CFD solution database for X-34, the "snapshot" approach of Proper Orthogonal Decomposition (POD) technique is applied to facilitate reduced-order modeling of steady state super- and hyper-sonic aerothermodynamics by varying either the angle of attack or free-stream Mach number. The scalar coefficients are further determined by the Response Surface Method (RSM) combined with Neural Networks (NN). For all Mach numbers under consideration, the first three POD modes suffice for a reasonably accurate reconstruction of the CFD solution over the entire range of angles of attack, suggesting a continuous variation of solution in the design space. The resulting s/hypersonic aerodynamic module is not only computationally more efficient than the low-level engineering methods but also as accurate as the high-level CFD methods. Such an aerodynamic module is best suited for multidisciplinary design and optimization. Copyright © 2005 by ZONA Technology, Inc.