Abstract: Genetic algorithm searches global optimum solution easily, but its local search optimization ability is poor and premature convergence and random roam can easily take place. Based on basic genetic algorithm theory and its practice course that this paper uses, some improved measures to its shortcomings are presented: by means of the chaos serial's properties of “ergodicity， randomness，regularity", original population is generated; the strategy that the best individual is saved and the worst individual is replaced is adopted. On the base of improved genetic algorithm, adaptive crossover and mutation rate formula are introdued. At the same time, the crossover rate and mutation rate are adjusted by extent coefficient to form adaptive genetic algorithm. By a numerical example of the fifteen bar truss, the optimal results and courses of the alyorithm are compared with that of basic genetic algorithm and improved genetic algorithm so that its advantage is demonstrated.

Key words: structural optimization, basic genetic algorithm, improved genetic algorithm, , discrete variables