Research On Ship Outfitting Design Experience Knowledge Retrieval Method Based On Knowledge Graph And Engineering Semantics

In order to meet the specifications in the design work,outfitting designers must thoroughly investigate relevant design knowledge and historical experience.Information overload,stemming from the large amount of experience knowledge generated by research and development(R&D)activities,reduces designers’ interest and efficiency in knowledge reuse.For an enterprise,the key to improving the efficiency and quality of knowledge reuse is helping designers understand and reuse existing ship outfitting design experience knowledge.Therefore,to improve knowledge utilization in ship outfitting design,this paper researches the management of ship outfitting design experience knowledge in the following aspects:First of all,to decease workload and reduce domain expert demand in ship outfitting design experience knowledge extraction,this paper proposes an automatic method of knowledge graph construction based on deep learning.The ship outfitting design experience knowledge graph was constructed from top to bottom,and the scheme of the ship outfitting design experience knowledge graph was designed.The automatic extraction based on deep learning was explored for the ship outfitting design experience knowledge.Entity boundary embedding,entity masking,and two-stage training were utilized to achieve improvement in extracting entities and relations from corpus.The feasibility and effectiveness of the proposed method are proved by comparing experiments.A ship outfitting design experience knowledge graph is constructed from the corpus.Secondly,the semantic matching in the ship outfitting design experience knowledge retrieval was researched.Considering the complexity of modeling and matching the engineering semantics in engineering experience knowledge retrieval,the knowledge hyper-network and the calculation of the semantic relevance are proposed.With the help of the domain knowledge graph and the deep learning model,the correlation between concepts and contents can calculate.The knowledge hyper-network of ship outfitting design quality cases is constructed to mine the complex engineering semantics of ship outfitting design experience knowledge.Based on Bayesian inference,the possibility of existing ship design quality case solving the new problem is predicted.Knowledge units and the user need are matched based on three-element dimensions,product,context,content,thus the accuracy of recommendation in knowledge retrieval is improved.Finally,an application,based on ship outfitting design experience knowledge retrieval and knowledge map,is developed to illustrate the proposed methodology.A specific case from a large shipbuilding company is shown to demonstrate the process of recommending existing outfitting design experience knowledge to designers in knowledge retrieval.