Knowledge Representation Combining Case-based Reasoning With Knowledge Graphs For Stamping Die Design

Product design depends on design knowledge,including all kinds of explicit design knowledge and implicit empirical design knowledge.Case-based reasoning(CBR)is an effective method to represent empirical design knowledge.However,the traditional CBR method lacks scalability and flexibility in the knowledge representation.In view of the huge demand of implicit empirical design knowledge representation and reuse and the issues of traditional CBR methods,this paper proposes a stamping die design knowledge representation method combining Case-based reasoning with knowledge graphs,which improves the efficiency and scalability of case representation and retrieval.This work has important theoretical significance and research value for better preservation and reuse of stamping die design knowledge.In this paper,the design of stamping dies and their design knowledge are deeply studied and analyzed.The process of stamping die design and design knowledge are discussed in detail.The advantages and disadvantages of traditional CBR technology are analyzed.An ontology model to represent stamping die design is proposed.Ontology modeling is carried out under protégé tool,which lays a foundation for the following knowledge representation of stamping die design.A novel method of retrieving design cases from knowledge graphs is proposed.This method uses the problem graph to represent the new design problem instead of the key words,and the design case retrieval problem is transformed into a semantic relation query problem.The query result is a subgraph of the design case base similar to the design problem query graph.In addition,this paper proposes a weighted similarity calculation method between a query graph and a similar graph.Through semantic similarity calculation and ranking for query results,the most similar design cases can be retrieved.Finally,three stamping die designs are selected as the analysis examples of design knowledge representation and retrieval.All related works are completed,including the construction of design cases,retrieval of candidate design cases,and similarity calculation and ranking,which shows that the proposed method is feasible and effective.The presented method is also compared with two traditional CBR methods used in two open literatures,which shows that the method proposed in this paper has more flexibility and scalability.