A hybrid optimization method using simulated annealing (SA) and a genetic algorithm (GA) developed in a previous study is improved in order to achieve high computational performance. The previous method, in which the genetic operations for generating initial search solutions of the next generation are introduced into the computation process of multi-point SA, can efficiently and globally compute Pareto-optimal solutions, as compared to SA using a re-annealing method. However, a large number of Pareto-optimal solutions exceeding the acceptable amount of computer memory may be obtained. In the proposed method, in order to effectively store Pareto-optimal solutions in computer memory, all obtained solutions are classified according to topology type, and those that should be stored are selected from each topology group. In addition, selection methods based on four criteria are newly proposed in order to improve the convergence of solutions and to maintain their diversity. The proposed SA/GA hybrid algorithm for simultaneous optimization of the topology and geometry of a two-dimensional framed structure is applied to a two-objective optimization problem of minimizing the total weight and maximizing the first natural frequency under displacement and stress constraints. The validity of the proposed methods is confirmed through performance evaluation based on the Wilcoxon rank-sum test.