Configuration synthesis of structure element plays an important role in middleware design and function block diagram design of control strategies. In order to overcome the defect-the low efficiency of classical optimization algorithm when combination scale is large, an improved particle swarm optimization (PSO) algorithm is proposed. Firstly cyclic natural number coding method is employed to describe the position and velocity of every particle. Then greedy strategy is adopted to convert the problem into a unimodal function optimization. Sectional inertia parameter and guaranteed convergence algorithm are used to improve the global convergence ability. The multiple optimal solutions are obtained by means of recording multiple global optimal positions in each time of iteration. Finally an example of 378000 combinations is given to verify the effectiveness of the algorithm presented. The instance shows that the algorithm is simpler than Niching Technique in multiple optimal values searching and better than basic PSO in global convergence ability.