Research On The Construction Of CNC Machine Tool Fault Knowledge Graph Based On Deep Learning

With the advent of the era of big data,how to discover and apply useful information in a large amount of exponentially rising data has become more and more the focus of research.Using images to describe complex text data enables people to grasp related information more accurately and quickly,and it is easier to understand the information,so knowledge graphs came into being.With the vigorous development of smart manufacturing,knowledge graphs have also attracted much attention in the field of manufacturing applications.Therefore,this thesis takes the Computerized Numerical Control(CNC)machine tool fault text in the field of manufacturing applications as the basic data,and studies the deep learning-based CNC machine tool fault knowledge map construction technology to effectively sort out and present relevant information units for users.Query relevant information through application functions to help users make reasonable decisions.In the stage of fault knowledge extraction of CNC machine tools,based on the entities and relationships in the domain defined by knowledge modeling,this thesis proposes a model for named entity recognition based on Bidirectional Encoder Representations from Transformers(BERT)model.First,the fault text of CNC machine tools is marked by the BIO marking method and the marked corpus is input into the BERT model to obtain the feature vector representation,and then the Bidirectional Long Short-Term Memory(Bi LSTM)is used to extract the contextual features.Entities are extracted through a Conditional Random Field(CRF)model.At the same time,this thesis also proposes an entity relationship extraction based on the BERT model for the fault text of CNC machine tools.First,the pre-processed data is pre-trained through BERT,and the generated text feature vector representation is input into the Bidirectional Gated Recurrent Unit(Bi GRU)to learn the context information features of the text,and then use the obtained feature representation as the input of the Attention model,analyze the importance of each component in the sentence,output the relationship category through the softmax function,complete the relationship extraction task,and finally get the entity relationship Triad.The experimental results show that the proposed entity recognition model and relation extraction model are feasible and effective.In the storage stage of CNC machine tool fault knowledge graph,this thesis imports the entity-relationship triples and structured data generated by unstructured data into Neo4 j graph database.At the same time,the common data between structured data and unstructured data is generated.The cosine similarity comparison of item faults is used for data fusion,and the visual display of CNC machine tool fault knowledge graph is realized through Neo4 j graph database,and the construction of CNC machine tool fault knowledge graph is completed.Through experiments,it is proved that the construction method of CNC machine tool fault knowledge graph developed in this thesis can extract key information from unstructured texts in the field of CNC machine tool faults,obtain the relationship between effective entities and entities,and finally combine structured texts to build a complete CNC machine tool fault knowledge graph,which has established a solid foundation for the subsequent specific applications of intelligent question and answer and auxiliary decision-making in the field of CNC machine tool failure.