Taking the example of traveling salesman problem, new heuristic strategy controlling the flight of partial dimensions in the swarm is put forward by analyzing the character existing in the edge-set intersection among particles, in order to resolve the problems of premature convergence and too-slow convergence existing in particle swarm optimization. The new strategy protects most edges belonging to global optimal solutions with high probability, inducts the flight of partial dimensions in particles, and cuts down the searching space of each particle, thus improves the searching efficiency of the population. The new particle swarm algorithm converges high-efficaciously to global optimal solutions of TSP whose scale are less than 2,000 cities.