Abstract: In view of that the standard genetic algorithm (SGA) has such frequent shortcomings as premature convergence, oscillation and over-randomization in iterative process, a genetic operator named transgenic operator is proposed to improve SGA. The transgenic operator can make good use of the information of which the adaptability has been calculated, keep the best individual from missing, and improve the adaptability of every chromosome in the population. The improved genetic algorithm (IGA) including the transgenic operator could be used directly to work out the optimum structural design with discrete variables to constrain both stress and crass-section area and could be dealt with the discrete structural optimization featured with multi-loading, multi-constraints and multi-variables. The results by exemplification show that the performance of convergence and optimization results of IGA are superior to that of SGA and it is an ideal method for optimum design of building structures.