Research On Knowledge-based Modular Gun Family Aided Design

Intelligent aided design of product family is an important development trend of product design methodology in the current engineering field,and modular guns have also become a hot spot in the development of guns in various countries.Based on the background of "modular gun family",aiming at the intellectualization and automation of gun family design,this paper studies the representation and acquisition technology of engineering knowledge,builds a knowledge base and knowledge-based aided design system for gun family design.The main contents include:Introducing the ideas and methods of product family design,it is proposed in this paper to divide the gun family design into two stages: the gun family product definition stage and the gun family module design stage.The tasks that need to be completed in the definition stage are to determine the specifications of the gun family products,divide the modules,establish the module composition scheme and interface scheme,and determine the main parameters of the gun family products;while in the module design stage,the task is to complete their selection,matching,parameter calculation,and finally the three-dimensional structure design according to the functional requirements of each module.Aiming at the problem of insufficient sample size and sparse sample density in the calculating process of gun family main parameters,Random Forest regression algorithm is introduced.Taking caliber,range,muzzle velocity,muzzle kinetic energy,maximum chamber pressure,etc.as input,barrel length and total gun weight as output,supervised training is performed on the calculation model,and estimated values of the main parameters are given in the design case.Based on the functional decomposition model of gun family,the design matrix is factorized based on the Binary Matrix Factorization algorithm and clustering analysis is conducted.A gun family module division and interface scheme is then formed combining the design knowledge of existing mature modular guns.The ontology conceptual model of gun family design knowledge is constructed,and the definition of entity type and relationship type is completed.Taking the ontology model as a formal and standardized reference,a knowledge graph of gun family design is constructed.The Bidirectional Encoder Representation from Transformers deep learning model is introduced to improve the efficiency of knowledge acquisition in the process of knowledge graph construction.Two QA systems respectively based on Machine Reading Comprehension technique and Knowledge-Based Question Answering technique are built.The experimental results of the two are compared and analyzed.On the basis of the knowledge graph,and in view of the diversification characteristics of design knowledge source and type,product data,design rules,design calculation process,engineering drawings and other knowledge types are respectively represented and stored to construct the general knowledge base for gun family design.The knowledge base provides data and knowledge support for the realization of knowledge-based aided design system.In the design process of the gun family modules,hybrid reasoning strategy of case-based reasoning,design rule and calculation cross-reasoning is used,and a hybrid reasoning procedure of modular gun family is formulated.In the case-based reasoning process,the concept of interface variance degree is introduced to calculate the overall matching evaluation value of the gun family design case.The cross-reasoning of design rule and calculation is used to check the reasonableness in the design process.Design cases of several main modules are used to exemplify the hybrid reasoning method of the modular design,and finally a 3D assembly model of the gun family is generated.According to the methods and strategies proposed above,the construction of the prototype gun family aided design system has been completed.The system is based on the general knowledge base,and is constructed using the Django website framework.Main parameter calculation,function decomposition and module division,module parameter design,knowledge graph visualization,knowledge retrieval,QA system,various database management and other functional modules are included in the system.Through the verification of some application cases,the system can preliminarily realize the intelligent solution in each design stage,and assist the designer in design decisions in the gun family design process.