Research On Intelligent Technical Forecasting Method Based On Evolution Route And Convolutional Neural Network

Under the background of increasingly fierce international competition environment and economic globalization,the independent innovation ability and core technology breakthrough of the country and even the enterprise are particularly important.As a means to comprehensively analyze the current status of technology development and predict the future potential technology field and direction of technology breakthrough,technical forecasting has been paid more and more attention by the state and enterprises.The evolution tree is a technical forecasting tool based on TRIZ theory.The construction of the evolution tree can help enterprises predict the potential development direction of technology and reasonably plan the design scheme of new product research and development,so as to enhance the comprehensive strength of enterprises.Therefore,it is of great significance for the research on the application of evolution tree.The first step in building an evolution tree is to determine the composition of the evolution route.At present,the identification of evolution route mainly relies on the artificial identification method of researchers,which is time-consuming and difficult to connect the research results of different people.Based on this recognition status,this paper makes a thorough study of the features of evolution routes and proposes a unified method of artificial identification of evolution routes and an intelligent identification method of evolution routes based on convolutional neural network.The main content of this paper includes the following three aspects:1.Summarize the composition of evolution route,construction process and application method of evolution tree.The general evolutionary features of the technical system along the evolution trend of the evolution route are summarized by studying the evolution nodes of the evolution route.By calculating the occurrence frequency of evolution features in evolution route examples,the primary evolutionary feature and secondary evolutionary feature of evolution route are analyzed on the basis of general evolutionary features and the evolution routes are classified and designed.Based on the characteristic research of evolution route,a method of artificial identification of evolution route is designed.2.The standard function set summarized through the functional expression of TRIZ theory standardizes the functional representation of the evolution tree technology system.A patent literature collection and screening method for technical system function is proposed.The paper summarizes the structured and unstructured characteristics of patents and determines the position of the patent information reflecting the evolution characteristics in the patent literature.Based on the direct observation and comparison of the two types of patent characteristics,the corresponding patent information model is established.3.The structural composition and three structural characteristics of convolutional neural network are introduced in order.The construction process of convolutional neural network for evolution route recognition is designed,which mainly includes defining the transformation relationship between the patent information and the input layer image and determining the super parameters of each layer of the network.A specific convolutional neural network was constructed by taking the identification of three evolution routes,namely single-double-multisystem evolution route,system cutting route and system expansion-cutting route.And the parameter training and recognition accuracy test of the network were completed.Then,the network structure and the setting of super parameters were optimized according to the test results.