关键词: 研发网络, 自适应, 断边重连, SIS, 风险传播

Abstract: In the modern society, faced with the complexity of technology, scarcity of the dynamic of market demands, it has become an effective mode for R&D project to form a network composed of many companies. Certainly, R&D network will help R&D firms to reduce costs and risks. However, when a firm in the network is not efficiently involved in the R&D project due to changes of the external market environment or its own poor management, this kind of event will have a negative effect on other firms that cooperate with him, which may propagate and result in paralysis of the network, this phenomenon is the cascading failure of the R&D network.
According to literature review, many studies focus on analyzing the cascading failure of the R&D network from the perspective of static network, few studies research the cascading failure of the R&D network from the perspective of adaptive network. Compared with the static network, adaptive network concerns that the network can actively change its topology during the cascading failure process. This is a ubiquitous and very important phenomenon, which can help us further explain the law of the cascading failure of the R&D network.
Firstly, the BBV algorithm is proposed to build the R&D network. And the roles are generated to describe the adaptive evolution process when a node is infected. The adaptive model consists of two steps, one is to choose the tie to break, the other is to choose the node to generate new tie for the susceptible node in the broken tie.
Secondly, the cascading failure model of the R&D network is proposed, combined with the characteristics of the R&D network and the SIS model. Compared with other cascading failure models, such as the Load-capacity model and the Mean-field, this model can represent the discrete spatial and temporal evolution of the cascading failure of the R&D network.
Finally, different effectiveness of these rewiring strategies is analyzed against the cascading failures of the R&D network under different values of some critical parameters. The simulation results show that (1) C1 strategy enhances the hierarchy and community strength of the R&D network and mitigates the propagation of risk in the R&D network to a certain extent. The establishment of new connections between nodes under C2 strategy is basically based on the proximity, which is easy to fall into path dependence and ability traps. (2) The adaptive behavior of R&D network will lead to the fluctuation of community strength, the decrease of average path length and the increase of average clustering coefficient reflect the effectiveness of C1 strategy. (3) Under the C1 strategy, there is a "U" correlation between the rewiring probability p and I*, and I* decreases gradually with the rewiring probability p under the C2 strategy. (4) Under the C1 strategy and the C2 strategy, as the parameter ζ increases, I* also increases. It is known that the level of organizational dependence is a key factor in the mitigation of the risk propagation of the R&D network. The rules of risk propagation of R&D network when considering the adaptive behaviors are revealed and a theoretical basis for R&D network governance in the context of network operation is provided.

Key words: R&D network, adaptive, rewiring, SIS, risk propagation