Application Research Of Improved GA-SA In Impeller Optimum Design

Simulated the natural evolution, Genetic Algorithm is a stochastic searching and optimizing algorithm, which has strong problem-solving abilities and extensive adaptability. It has been widely applied in the optimum design because of its robustness, flexibility and easy to operate.Impeller is the core component of high-speed rotating centrifugal compressor, which structure design is directly related to life length of the impeller, accuracy of creating a sine wave and electricity consumption, and so on. Therefore, in the design of modern high performance compressor, there is great significance on structure optimum design of the centrifugal impeller.On the one hand, genetic algorithm has been widely used because of its easy operation and solving the problem effectively.On the other hand, such as premature convergence, local search capabilities are still the problems of genetic algorithm in practical applications. This article focuses on the search of improving the genetic algorithm, and its application in the impeller's optimum design.This article has some researches and analyses about the theory, optimization and application of genetic algorithm. Firstly, we focused on analyzing the shortcomings of GA, and sought some solutions. Considering the randomicity which the simple genetic algorithm produces the initial population, we designed a new method of producing the initial population. It uses the max-min distance means and makes various individuals maintain some certain Hamming distances to produce good population distribution. Secondly, the article introduces some improved methods, such as: Adaptive Genetic Algorithm, Hybrid Genetic Algorithm, and so on. A new adaptive genetic algorithm is proposed on the basic of the existing adaptive genetic algorithm to improve the crossover probability and mutation probability. It bases on the population diversity and individual fitness to adjust the crossover probability and mutation probability automatically. Simultaneously, simulated annealing mechanism is introduced to do local-search and stretch fitness value suitable for the best chromosome in every generation of the evolution process. Some experiments show that the proposed new algorithm has a better improvement in global convergence and stability and gets the expectation effect.Finally, this improved algorithm is applied to the optimum design of impeller, and the simulation reveals its reliability.