Research On Knowledge Fusion And Reasoning Technology Oriented To Manufacturing

With the vigorous development of the Internet of Things,digital technology and artificial intelligence,knowledge fusion and inference technology are becoming more and more important in the application of knowledge engineering.The higher word vector dimension results in lower algorithm efficiency and higher operation cost.Secondly,multi-person collaborative modeling leads to high fragmentation,ambiguity and redundancy of ontology models,and lacks effective fusion methods.In addition,based on the constructed ontology model,how to accurately launch knowledge reasoning based on the search information to provide accurate knowledge results for the knowledge push end is an important part of the knowledge application task.In response to the above-mentioned problems,this paper carries out the following research work:In order to improve the operation efficiency of the algorithm,reduce the training cost,and ensure the efficient execution of downstream tasks,the dimensionality reduction method of semantic word vectors in the field of manufacturing knowledge is studied.This topic crawls the corpus of manufacturing knowledge and uses Word2 vec training to generate 50-dimensional and 200-dimensional word vectors.By analyzing the accuracy and training cost of word vectors,the necessity of second dimensionality reduction of word vectors is studied.The improved stacking autoencoder and PCA two dimensionality reduction methods are used to reduce the dimension of 200-dimensional vectors.The comparison experiment of semantic correlation and training cost for the same dimension of the same word is used to prove the effectiveness of the improved method.In order to solve the problems of fragmentation,ambiguity and high redundancy of ontology models in the process of multi-person collaborative modeling,the global similarity of manufacturing knowledge ontology and its fusion method are studied.Construct an entity similarity model to study the similarity between nodes from an entity perspective;propose a maximum path subgraph model to study the similarity of fragment ontologies based on path s tructure from the perspective of path structure;define the global similarity of fragment ontology based on the above two models.The type of knowledge ontology of fragment manufacturing is studied,and the entity disambiguation rules and logic structure fusion rules are proposed to achieve efficient and intelligent fusion of fragment ontologies.A plant ontology model was used to verify the feasibility of this method,and the integration of fragment ontology to complete manufacturing knowledge ontology was completed.In order to obtain more accurate and comprehensive knowledge push results,study knowledge inference technology based on manufacturing knowledge ontology.Using protégé to build and visualize manufacturing knowledge,establish a manufacturing knowledge base,realize the ontology management of manufacturing knowledge documents,and complete the connection between ontology models and knowledge documents.Research the location expansion method based on search keywords,propose a concept matching inference algorithm,and complete the knowledge inference process.Develop a knowledge push system for manufacturing knowledge,and verify the accuracy of knowledge inference algorithms according to evaluation rules.