Abstract: According to the research paradigm of network modeling from the perspective of system science， the main body， resources and dynamic mechanism of the knowledge activities in complex product industrial clusters are analyzed. A super network model of multi-organization knowledge learning with dynamic characteristics is constructed. It is found by going deep into the evolution process of the model through simulation experiments that the organizing characteristics of different project teams will bring different industrial clusters' learning performance and there exists an optimal project team size that can achieve maximum performance. When the topological structure of the organizational sub-network where the project team located is in the small world characteristic， the knowledge level of multi-ganization knowledge learning super network can rise rapidly， and learning performance is high. When the project team size is smaller than the optimal scale， the duration of the small world structure of the organizational sub-network is inversely proportional to the project team scale， etc. There are still some works to improve the learning performance of the complex product industrial cluster， including optimization of the public knowledge and information platform， control of the size and structure characteristics of the project team， and management of the multi-organization learning network， etc.

Key words: super network modeling, complex product industrial cluster, knowledge management, learning performance