Abstract: In order to realize the resource sharing and risk sharing during research and development, enterprises have joined in the technological innovation network to share and exchange resources, and jointly develop new products. But in reality, the technological innovation network is always in a state of dynamic evolution under the influence of various factors in industrial life cycle. This instability seriously weakens the innovation performance of the network, which not only affects the structure and function of the technological innovation network, but also affects the behavior of each individual in the network, and ultimately affects the success of technological innovation. Exploring evolutionary motivation of technological innovation network is helpful to network reconfiguration and performance improvement. Although the current research has achieved some positive results, the evolution of innovation network is not only reflected in the formation of new relationship, but also in the dissolution of old relationship. When the target node is invalid or inaction, it will be moved from the cooperative relationship, or even eliminated by the network. The research on the factors of relationship dissolution is insufficient, and it is not clear whether the factors driving the relationship formation and the dissolution are consistent. Based on the life cycle theory, the factors that dominate the evolution of technological innovation network are different at different stages of industrial development. It is also necessary to further explore the factors that play a role in the formation of relationships and the dissolution of relationships at different stages of industrial development. This paper makes up for the shortcoming of previous research and constructs an evolution dynamics model of technological innovation network based on TERGMs (Temporal Exponential Random Graph Models) by integrating firm exogenous covariates and endogenous structural variables, distinguishing the process of network relationship formation and dissolution, then empirically studies the technological innovation network of OLED (Organic Light Emitting Diode display technology) from the perspective of life cycle. Specifically, it combines the exogenous covariates (technological innovation ability and homogeneity) and endogenous structure variables (degree, edge-sharing-partner, degree-sharing-partner), and puts forward four research hypotheses: the relationship between exogenous node attribute and cooperation formation, the relationship between exogenous node attribute and cooperation dissolution, the relationship between endogenous structure variables and cooperation formation, the relation between endogenous structure variables and cooperation dissolution. The research hypotheses are as follows:H1a: In the evolution of innovation networks, individuals with strong technological innovation ability are more likely to build cooperative relationships.H1b: In the evolution of innovation networks, individuals with similar technological innovation ability are more likely to build cooperative relationships.H2a: In the evolution of innovation networks, individuals with strong technological innovation ability are more likely to avoid dissolving the cooperative relationships.H2b: In the evolution of innovation networks, individuals with similar technological innovation capability are more likely to dissolve the current cooperative partnerships.H3a: In the evolution of innovation networks, individuals with high degree are more likely to avoid building new cooperative partnerships.H3b: In the evolution of innovation networks, individuals with edge-sharing-partner are more likely to build new cooperative relationships.H3c: In the evolution of innovation networks, individuals with degree-sharing-partner are more likely to build new cooperative relationships.H4a: In the evolution of innovation networks, individuals with high degree are more likely to avoid dissolving current cooperative relationships.H4b: In the evolution of innovation networks, individuals with edge-sharing-partner are more likely to avoid dissolving current cooperative relationships.H4c: In the evolution of innovation networks, individuals with degree-sharing-partner are more likely to avoid dissolving current cooperative relationship.Then a progressive model of the evolution dynamics of technological innovation network is constructed. Using the 20-year patent application data of OLED industry from 1995 to 2014 in Derwent Innovation Index, the evolution of OLED technology innovation network is divided into four stages: the incubation stage (1995-1999), the growth stage (2000-2004), the mature stage (2005-2009), and the regeneration stage (2010-2014) according to the S-Curve of patent application, then this paper makes an empirical study on the dynamic factors of evolution on OLED technological innovation network, in which the technological innovation network is constructed by the patentee co-occurrence data from patent application data. The network structure changes significantly in different life stages: the network structure in the mature period and regeneration period presents a more complex network relationship, and finally realized network connectivity. Throughout the life cycle, the increasing network density means that the members are more and more closely connected, and the network is more and more clustered, in addition to the growth stage. The steady increase in the number of patent applications shows that technology can more and more meet the market demand, while a good market is conducive to the better development of technology, then technology and market continue to promote each other to achieve stable development.The empirical results of OLED technological innovation network show that on the one hand—about the formation of relationships, the technological innovation ability of enterprises plays a significant positive role in the incubation period and growth period, while the edge-sharing-partner shows a significant positive effect in the growth period, and the degree and degree-sharing-partner have a significant negative role in the mature period and regeneration period. On the other hand—about the dissolution of relationships, degree and degree-sharing-partner have a significant negative role in the mature and regeneration period, while other factors did not show significant effect. The results confirm that in different stages of industrial development, the factors affecting the evolution of technological innovation network are different, and the factors that affect the formation of relationship and the dissolution are also different. Finally, the validity of the proposed model is verified by MCMC diagnosis and good-of-fit test. It is proved that the model is a stable and effective model which can well describe the observation network, and the parameters included can represent the structural characteristics of the network.The main contribution of this study are as follows: in theory, it distinguishes the relationship formation and the dissolution process of technological innovation network, and puts forward the model of studying the evolution of innovation network, which is applicable to the analysis of the evolutionary dynamics of different innovation network; in practice, the empirical results of the evolutionary dynamics of OLED technological innovation network are helpful in understanding the current situation of industrial innovation network, guiding industrial innovation and realizing scientific governance.

Key words: technological innovation network, evolution dynamics, TERGMs, OLED