Manufacturing Quality Formation And The Key Process Evaluation For The Parts Of Nuclear Power Equipment Based On Knowledge Graph

Intelligent manufacturing is the main direction of manufacturing industry development,which is based on the background of new generation of information technology,and is the product of the deep integration of manufacturing technology and information technology.Among them,the intelligent manufacturing quality management is one of the main directions of intelligent manufacturing development.As national strategic basic equipment,the intelligent level of manufacturing quality management of nuclear power equipment reflects the overall technical level of China’s manufacturing industry.However,at present,a lot of nuclear power equipment manufacturing quality related data stored in the form of a traditional relational database or text,which makes the manufacturing quality formation face issues,such as,data isolation and weak correlation,high redundancy and the text information,and leads to the difficulties in data mining and the use of hidden information.It is a great challenge to promote the intelligent development of nuclear power equipment manufacturing quality formation process.Therefore,this thesis firstly analyzes the formation process of manufacturing quality for the parts of nuclear power equipment,and defines the prior knowledge and process knowledge of manufacturing quality.Secondly,the ontology model of manufacturing quality formation was constructed to establish the relationship of manufacturing process quality knowledge on the semantic level.At the same time,the knowledge graph of manufacturing quality formation was generated by constructing the extraction framework of manufacturing quality knowledge,and the intelligent brain was established for manufacturing quality formation process.Finally,a method for evaluating the importance of process was proposed based on knowledge graph,which realized the intelligent control of manufacturing process quality and provided a new idea for a new generation of quality management mode.The main research contents of this thesis are as follows:Firstly,an ontology modeling method of manufacturing quality formation based on the knowledge tree of manufacturing quality was proposed to construct the ontology model of manufacturing quality formation for the parts of nuclear power equipment.Firstly,the quality knowledge related to manufacturing process is analyzed,and the quality knowledge involved in product design stage and the quality knowledge generated in manufacturing stage are defined as quality prior knowledge and process knowledge.Then,a manufacturing process quality knowledge structure tree model was designed to realize the hierarchical division of manufacturing process quality knowledge.Finally,the ontology model of manufacturing quality formation was constructed by using the structure tree,and the semantic correlation between quality knowledge in design stage and manufacturing stage was realized.Secondly,a method of knowledge extraction,fusion and storage for manufacturing quality for the parts of nuclear power equipment was proposed in terms of knowledge graph generation.Firstly,the word2 vec pretraining model is used to represent unstructured quality knowledge as word vector,so as to convert text characters into numerical values,thus speeding up the training speed of subsequent models.Then,the architecture of unstructured knowledge extraction based on Bi LSTM-CRF network structure and relation word trigger is established to realize named entity recognition and relation extraction of unstructured quality knowledge.At the same time,the transformation of structured quality knowledge(relational database)to triplet data is realized by using the rule mapping table developed by R2 RML language.Finally,in order to ensure the consistency of the obtained quality knowledge,a knowledge fusion method based on Jaccard coefficient and Levenshtein minimum edit distance was constructed,and the triad data was saved in the Neo4 j graphics database to generate the knowledge graph of manufacturing quality formation.Thirdly,in the application of knowledge graph of manufacturing quality formation for the parts of nuclear power equipment,an evaluation method for evaluating the key processes based on knowledge graph of manufacturing quality formation was established.Firstly,according to the difference of the influence degree of each manufacturing quality factor on the quality of each process,the quality factors in the manufacturing process are divided into different grades to determine the weight distribution in the calculation process of Page Rank algorithm.Then,the equipment reliability factor is embedded to characterize the influence of quality on the equipment degradation process,and the improved Page Rank algorithm is obtained.Finally,the activity degree of each node in the knowledge graph of manufacturing quality formation was determined by using the improved Page Rank algorithm,and then the key processes in manufacturing process were evaluated through ranking the activity degree of each node.Fourthly,taking the blade manufacturing quality forming process of nuclear power turbine unit as a case study,the knowledge graph of blade manufacturing quality formation was generated,and the importance of processes of blade manufacturing process were evaluated by using the graph.First,an ontology model of blade manufacturing quality formation was constructed.Then,in order to verify the effectiveness of the knowledge extraction framework,the knowledge extraction effects of BILSTM-CRF and LSTM and Bi LSTM are compared and analyzed.Meanwhile,the knowledge graph of blade manufacturing quality formation was generated by using the knowledge extraction framework.Secondly,the PR values of each process node were determined by using the improved Page Rank algorithm(IPR),and then the importance of processes in the blade manufacturing process were evaluated.The validity of IPR algorithm is verified by analyzing the agreement degree between the identification results and the importance of processes summarized from experience,and the rationality of knowledge graph method is also explained at the same time.Finally,the advantages of IPR is verified by comparing with PR and WPR algorithms.