Methods of Uncertainty Propagation for the Conceptual mission integrated design optimization of a hypersonic aircraft

Dominic Joseph Agentis (Virginia Polytechnic Institute and State University)

CHALLENGE - The focus of this work is to capture the effects of uncertainties on the design optimizations of a hypersonic aircraft completing a given flight mission in which the values of selected parameters are inexact and to assess the effectiveness of the methods used to capture the uncertainties.

SOLUTION - A set of single-objective system design optimizations is conducted on the model to compare results when using either the Gaussian Process Regression  or Polynomial Chaos (PC) uncertainty propagation methods. An optimization is comprised of a defined series of design iterations, each with a unique combination of input variable values that correspond to a given design of the hypersonic vehicle. A full mission simulation is run for each design using the given input decision variable values for that iteration. The optimization algorithm analyzes the simulated results from previous iterations and schedules the input decision variable values for following iterations with the goal of maximizing or minimizing the value of an objective function defined by a simulation output. modeFRONTIER optimization software is employed to execute the optimization algorithms in this study.

BENEFITS - The use of a multiple-strategy optimization algorithm combines the benefit of evaluating multi-dimensional design spaces found using an evolutionary optimization algorithm with the rapid convergence to an optimum once a feasible design is discovered using a gradient-based optimization algorithm. The Hybrid method also allows for parallel evaluation of design configurations that allows for greater exploration of the design space in a given amount of time relative to the computational resources available.