An Improvement of Genetic Algorithm and Its Application to Aerodynamic Design
Fan Huiyuan   Wang Shangjin   Xi Guang
(Xi'an Jiaotong University, 710049)

Abstract:  An idea of a probabilistic binary search is proposed to improve the standard genetic algorithm (SGA) by adding a step of probabilistic binary decision in the SGA procedure. In the added step, the fitness for each components of chromosomes(binary strings) in the population is first calculated, and then is used to count statistically the values at the component's positions (i.e. genic position) during the past evolution, and each component's position will be assigned with a score. Finally, based on these scores, several new individuals were generated and joined to the population in next generation to be evolved. Due to the inclusion of some global evolution information in the scores, it is expected that so-obtained individuals have "better" qualities, and can make the algorithm improving convergence. The improved genetic algorithm was verified with a maximum value searching problem of a 2-dimensional multimodal function. It was then applied to carry out the blade inverse design of a centrifugal compressor diffurer.The comparison between the result of the new approach with that of SGA demonstrates the new algorithm is superior to the SGA for solving the chosen problem.
Keywords: genetic algorithm, binary search, optimization, aerodynamic design.