Innovation diffusion is an important driving force to technology upgrade and maintaining the competitive advantage for industrial clusters. Based on complex network theory and innovation diffusion, this paper establishes an industrial cluster evolution network model, uses neighbor infection weight method, and puts forward an modified SIR epidemic model which has three infected stage. It simulates the innovation diffusion process in industry cluster network by computer simulation. Simulation results shows that the strength of the nodes has an important influence on innovation diffusion, the adjustable parameter K has a positive correlation with the innovation diffusion, and there is a threshold value for the innovation diffusion of industrial clusters. Finally, some policy implications are put forward. The article has a certain academic reference value for the related research.