Construction Of Cement Clinker Quality Knowledge Graph

With the development of "Intelligence+" in the industrial internet,the digital transformation of the cement industry is actively underway.Knowledge graph,with its outstanding ability in knowledge storage,expression,and reasoning,effectively solves the problem of independent knowledge distribution and reliance on human experience in the cement industry.Therefore,constructing a knowledge graph in the field of cement clinker quality and applying it to the cement production process is of great significance for improving the quality and production efficiency of cement clinker.Currently,many scholars have conducted in-depth research on the construction and expression of knowledge graphs in general domains,but there is little research on the construction of knowledge graphs in the field of cement clinker quality,and the important auxiliary role of entity attributes in knowledge reasoning or decision-making in the field of cement clinker quality has also been neglected.Therefore,the research content of this paper is the construction of a knowledge graph of cement clinker quality,and the specific contents are as follows:(1)Establishing a corpus in the field of cement clinker quality.Based on the cement production process,the composition of clinker calcination and raw materials is closely related to the quality of cement clinker.Knowledge data is obtained from CNKI,technical manuals,and operation documents of cement enterprises.The obtained knowledge data is then cleaned and preprocessed,and a reasonable labeling scheme is customized for BIOES sequence labeling to pave the way for entity recognition and relationship extraction in the later stage.(2)Entity recognition in the field of cement clinker quality.A combined Ro BERTa and Bi LSTM-CRF-WOL model is designed for entity recognition.The Ro BERTa model can handle the diversity of entities in the field of cement clinker quality and extract deep-level semantics.Bi LSTM-CRF is used for sequence labeling and decoding of the semantic encoding output of the previous layer.The model introduces loss weight to reward or punish the entity recognition results,improving the performance of the model.The effectiveness of this entity recognition model is verified through comparative experiments with several other models.Finally,entity attributes are identified through dependency syntax analysis based on entity recognition.(3)Relationship extraction in the field of cement clinker quality.A Ro BERTa-based entity relationship extraction model called Ro BERTa-PCNN-Attention is designed for the characteristics of long texts and polysemy in the corpus of cement clinker quality.The Ro BERTa model can better understand the context information of sentences,and the self-attention mechanism assigns different weights to each element in the sentence,extracting the required triplets for the knowledge graph by comprehensively considering the overall features of the sentence.Comparative experiments with other models verify the good relationship extraction capability of this model.(4)Construction of a knowledge graph of cement clinker quality.First,knowledge fusion is conducted for problems such as synonymous and polysemous words in the extracted entities and their relationships.Then,the Neo4 j graph database is selected for knowledge storage through comparison with other databases,and finally,a knowledge platform is designed on the basis of the graph to assist cement enterprises in their production work.