Feature-Based Knowledge Base Building And Data Mining For Aeronautical Structural Parts

With the continuous development of aerospace manufacturing,the structural parts have been widely used because of their good performance,and the number of enterprises knowledge is also increasing.Knowledge has the characteristics of complexity and empiricism.Due to the lack of effective collation and induction,the reusability of enterprise knowledge is reduced.The variety of structural parts and the tedious technology make it difficult to process parts,and depend more on expert experience in process design,which leads to the longer cycle of parts process design.Therefore,it is of great significance to study knowledge effective organization and expression and excavate valuable knowledge which can provide support for process design.In this paper,the aeronautical structural parts was taken as the research object to study the classification and expression methods of process knowledge.Firstly,the method of knowledge division was studied.And the classification of the structural parts process knowledge was determined based on the actual production and processing of the company.Then the corresponding expressions were established on the basis of different classifications.Finally,the structure of knowledge base which conformed to the knowledge expression form was constructed and laid the foundation for the subsequent establishment of a process knowledge base system.The development of the process knowledge base system needed to clarify the data requirements and functional requirements of the system firstly,and design the system function structure diagram and the system workflow diagram.Secondly,on the basis of three kinds of knowledge classification and system requirements analysis,the architecture of the aerospace structural process knowledge base system was established combined with the conceptual model design and logical model design method,and the logical model design results were stored through the SQL Server 2008 R2 database.Finally,in the Visual Studio 2010 development environment,the C#programming language was used to develop the system which realized the query and maintenance functions of process plan,cutting parameters,tool and machine tool information,and the functions of system backup restore and dynamic expansion of process terminology.It provided platform support for process design of parts.In order to obtain feature-based optimization tools,the part information,processing conditions and tool properties were analyzed and the factors that have a small impact were eliminated.And tool matching rules based on part information and machining conditions were established.The hierarchical structure model of tool was studied by Analytic Hierarchy Process(AHP),and the weight distribution of each tool factor was studied by combining the binary contrast method and square root method.The experts participated in the evaluation of the primary tool set.And the tools were optimized based on the fuzzy mathematics theory and the principle of maximum degree of membership.It provided the basis for the craftsman to select efficient tools.Feature-based prediction of cutting parameters was researched.The influence factors of the cutting parameters were analyzed,and the input layer,the hidden layer and the output layer node number of the BP neural network were designed.On the basis of the transfer function,training function,learning function and performance function,the network model was trained and the generalization ability of the network model was verified.The process plan of aeronautical structural was taken as an example to carry out mining research.First of all,the code structure was used to encode the process and the work step.Secondly,the overall similarity calculation of the process scheme was studied based on the multi-level similarity comprehensive measure method and the Kuhn-Mnkres algorithm.And the nearest neighbor propagation algorithm was used for clustering mining of typical process.Finally,the evaluation index was used to judge the effectiveness of the clustering results and provide a typical process plan for the process design.