Preparation of Advanced CFD Codes for use in Sensitivity Analyses and Multidisiplinary Design Optimization

Abstract

Several topics concerning computational methodologies relevant to preparation and efficient use of advanced CFD codes in sensitivity analyses (SA) and multidisciplinary analysis and design optimization MDO) are briefly discussed. This work was initiated several years ago by NASA Langley and has continued to involve government and university personnel. The several topics to be discussed and illustrated with sample results are as follow: Efficient calculation of both geometric and nongeometric sensitivity derivatives (SD) from a 2-D thin layer Navier-Stokes code and a 3-D supersonic (marching) Euler code using incremental iterative methods. Automatic differentiation (AD) of a 3-D thin layer Navier-Stokes code to obtain: geometric and nongeometric SD typical of those needed for MDO CFD algorithm and turbulence modeling SD, illustrating feasibility for SA of complex math models. Initial application of AD to incremental iterative forms for both first and second derivatives for a 2-D thin layer Navier-Stokes code. Development of a method for Simultaneous Aerodynamic Analysis and Design Optimization, SAADO. Some of these techniques have already been suggested for doing parts of the problem more efficiently; the contribution here is demonstrating how they fit together in a mutually beneficial way.