Self-Organising Method and Co-Evolutionary Mechanism of Collaborative Networks in Social Manufacturing

With the rapid growth of collaborative production demands, social manufacturing (SM) has emerged as a decentralized paradigm that enables the dynamic allocation of socialized manufacturing resources (MRs). Organizing manufacturing cluster into collaborative networks under this paradigm is vital for the aggregation of discrete resources. However, due to the self-interested nature of manufacturing entities, traditional network formulation methods fail to mitigate the risks of structural paralysis caused by passive behaviors. This paper proposes a dynamic collaborative network generation and evolution mechanism based on multi-agent evolutionary games. By quantifying the collaborative willingness of participants, this mechanism resolves the difficulties of precise resource clustering and stable network evolution. Specifically, a collaborative potential (ColP) model is established by integrating the attraction of capability complementarity, the repulsion of spatial distance, and dynamic load constraints to drive the self-organized generation of network topology. Next, a multi-agent game model is designed, where the connectivity and trust states of agents are mapped as penalty costs. Finally, a trust-driven dynamic rewiring mechanism is proposed for network evolution, employing the Fermi rule to update the strategies of agents. Experimental results demonstrate that the proposed models significantly improve the modularity and resource alignment of the collaborative network compared with benchmark paradigms.