Research On Marine Diesel Engine Fault Diagnosis Technology Driven By Knowledge Graph

As an important research direction in the field of marine diesel engine fault diagnosis,how to use artificial intelligence technology to accurately and efficiently make judgments on various abnormal conditions of marine diesel engines and reduce the fault loss to a minimum level is one of the important challenges in the development of the whole shipping industry in the direction of multi-dimensional intelligence.Traditional diagnostic methods are cumbersome to implement or are not efficient enough,making it difficult to manage diagnostic data and failing to be better transformed for processing and mining.In order to effectively solve this problem and realize the knowledge modeling and application management in the field of fault diagnosis,this thesis combines the characteristics and applicability of knowledge graph and deep learning to build domain knowledge graph and classification update technology for marine diesel engine fault cases,then develop a fault diagnosis platform for marine diesel engine with the guidance of this technology to achieve fault alarm information analysis and fault diagnosis assisted decision.The specific research contents are as follows:(1)Overall technical solution design for marine diesel engine fault diagnosis platform.By estimating the possible technical means,development tools and environment,applying the professional design concept and layout of development ideas,summarizing the existing diagnostic business processes and problems in the scheme,the overall scheme of marine diesel engine fault diagnosis is formulated,and the functional modules of the platform are divided and described.(2)Research on knowledge graph construction techniques in the field of fault diagnosis.Firstly,the knowledge cases of marine diesel engine fault diagnosis are collected and analysed,and the extracted conceptual information is categorised and modelled,and the knowledge graph is constructed using the collaborative approach of the knowledge graph schema layer and data layer;Secondly,ontology construction by mining the concepts and logic in the knowledge pre-proposal,leading to pattern layer construction;Finally,the data layer is constructed based on the pattern layer using deep learning methods,and the structured knowledge representation is outputted and stored in the graph database through knowledge extraction of the knowledge precisions,thus completing the construction of the knowledge graph.(3)Research on fault diagnosis knowledge graph classification update technique.Given that the constructed knowledge graph needs to be regularly maintained to keep up with updates in domain expertise,this thesis builds a new graph transfer learning network model based on graph convolutional neural networks that can learn class discriminative node representations with given source and target network label information,realizes the transfer of knowledge from partially labeled source attribute networks to assist the classification update of the knowledge graph,and completing the layout of a knowledge graph maintenance technology scheme in conjunction with this technology for optimising the analysis and judgement of fault information.(4)Marine diesel engine fault diagnosis platform development.With the support of the above technical achievements,combined with the fault diagnosis technology management process,the marine diesel engine fault diagnosis platform was developed,the working mechanism and operation procedures of each functional module were calibrated in turn,and the feasibility of the platform was confirmed by actual cases.