By depicting the rational innovation cooperation decision making behaviors of industry cluster members, the industry cluster innovation cooperation network evolution model is built, and then the network selforganizing evolution rules and its topological structures etc are analyzed by simulation methods. The results show that the higher the knowledge spillover efficiency is, the faster the network self organized to the stable state and the higher the network resource allocation efficiency and its fairness is. When the knowledge is partially spilled, the influence powers of cooperated members are not assertive and the emergent networks have common smallworld characteristics. The complementary knowledge is not the necessary condition for establishing the cooperation relationships, and the network position of potential partners is also an important influence factor which is independent of the knowledge levels.