Research On Process Reasoning And Optimization Method Based On Knowledge Graph

There are a large number of heterogeneous multi-source data in the machining field of mechanical parts,which are independent from each other and difficult to form a unified organic whole.As a result,designers cannot deeply use the existing data to guide the process reasoning and optimization of mechanical parts,which reduces the utilization rate of knowledge.Aiming at the above problems,this paper introduces knowledge graph technology to conduct knowledge modeling and data management for data in the field of mechanical parts,and studies the application of knowledge graph in process reasoning and optimization.The main research achievements are as follows:(1)Construct knowledge graph based on mechanical part features.Firstly,the feasibility of constructing knowledge graph at the granularity of mechanical part features is analyzed and verified,and a top-down construction strategy is determined.In order to improve the ontology quality,the schema layer is constructed by combining the domain expert experience and the improved seven-step method.For the data layer,there are two types of data:text data and relational data,which need to be processed separately due to their large format differences.For text data,the BERT-BiLSTM-CRF model is used for entity recognition,and then the template-based relational extraction method is used to extract triples.For relational data,R2RML is used to dump it into equivalent triplet data.In order to reduce knowledge redundancy,text similarity algorithm is used to realize knowledge fusion.(2)Study the completion algorithm of the feature knowledge graph of mechanical parts.In order to ensure the integrity of knowledge graph,a RGAT-HoLE completion algorithm is proposed.The algorithm also considers the semantic information of nodes and relations in the knowledge graph,and can effectively extract the features of nodes and relations.Good results are obtained in the FB15k237 data set and the part feature knowledge graph constructed in this paper,which verifies the effectiveness of the algorithm.(3)Study process reasoning and optimization method based on knowledge graph.Mechanical parts process reasoning and optimization includes two tasks:parameter reasoning and process sequence optimization.For parameter reasoning problem,this paper proposed a feature selection algorithm based on K-order neighborhood retrieval of knowledge graph.By this algorithm,the attributes most relevant to the machining process were selected,and then the process parameter reasoning was realized by SSA-BP model.For the process sequence optimization problem,this paper combines the constraint matrix and cost function to obtain the process cost matrix,and transforms it into the asymmetric traveling salesman problem(ATSP)by adding virtual process nodes,and introduces the LKH algorithm to solve the process sequence with the optimal cost.Experiments show that LKH algorithm is superior to conventional genetic algorithm in terms of solving results and efficiency.(4)Develop the knowledge graph management system for machining of mechanical parts,which integrates the above achievements,can realize the management of multi-source heterogeneous data in the field of mechanical parts,and provide parts process reasoning and optimization services.