Speedups for Efficient Genetic Algorithms : Design optimization of low-boom supersonic jet using parallel GA and micro-GA with external memory

GAs have been successfully used in an aerodynamic shape design optimization. Time efficiency issues related to the evaluation of a fitness are becoming a critical point as large calculations are needed. In this paper two efficient methods are applied to the original GAs to save computational time. Firstly, parallelized GA is used for a single disciplinary optimization to investigate the scalability corresponding to the number of processors. Without other disciplines, only a minimization of sonic boom is sought using a parallelized GA. The minimization of the sonic boom ground signature often leads to the undesirable aerodynamic properties such as increase of the drag. These disciplines are often conflicting, and it is important to balance aerodynamic performance and sonic boom requirements in a way that represents the best compromise for the overall design. Therefore, secondly, multi-objective methodology can be used to find a set of non-dominated solutions called the optimal Pareto sets for minimizing drag and sonic boom at the same time. Non-dominated sorting GA is used to get these optimal Pareto sets. To decrease the increase of computation time caused by adding more disciplines and comparing the ranks each other, micro-GA with external memory is inserted in the original non-dominated sorting GA. Results of a shape optimization of a low-boom supersonic jet with 15 design variables are presented. CPU time in running a parallel GA and micro-GA with external memory is compared with a serial GA and a usual nondominated sorting GA respectively. Both methods are shown to be successful by decreasing run time significantly.