Aerospace Vehicle Concept Selection Using Parallel, Variable Fidelity Genetic Algorithms
Michael Buonanno, Dimitri N. Mavris · 10th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference · 2004
In recent years a large amount of work has been done by industry, academia, and government with the objective of improving preliminary aircraft design methods. These new techniques have helped to bring about a paradigm shift that allows designers to gain insight into the impact of the configuration parameters on system acceptability early in the design process before excessive resources have been committed. Despite these advances, little progress has been made in creating rigorous methods to aid in the first stages of conceptual design. Currently, the prevailing method is to use qualitative techniques to select a handful of designs from the billions possible and then perform detailed analysis on each of these concepts in order to determine its performance and characteristics. Other methods have also been introduced to help the decision maker choose which concept from that handful is most worthy of further investigation based upon criteria limits, but these instantly exclude the vast majority of the concept space based on the designer’s intuition or biases and may prevent the selection of the best concept for the design mission. This paper describes the development and application of a modified Genetic Algorithm to the aircraft concept selection problem, and gives results for the optimization of a notional small supersonic transport.