Research On Fault Diagnosis Method For Complex Industrial Process

The development of science and technology has made modern industrial production more and more intelligent and complicated.It is true that complex industrial process improves the economic benefits of enterprises,but due to the large scale,complex structure and strong coupling between production units of complex industrial process,it is more and more difficult to fault diagnosis of complex industrial process.Because the production process generates a large amount of data,in the complex industrial process fault diagnosis,the way based on data processing and knowledge is the mainstream development direction.However,the influencing factors of complex industrial process have increased,and there are intricate associations among various influencing factors.The data-driven approach cannot well express these complex relationships,and the knowledge-based approach has the problem of efficiency in dealing with large-scale data,so the traditional fault diagnosis method cannot well adapt to the complex industrial process fault diagnosis scenario.Therefore,the fault diagnosis technology of complex industrial process has become a problem worth studying in recent years.In response to the above problems,we need to adopt a novel,general and complete solution to solve the difficulties encountered in the diagnosis of complex industrial process.Based on this,this thesis uses a method based on the combination of knowledge and data drive to conduct research.The main contents of this article are as follows:(1)This thesis proposes a fault diagnosis scheme based on multi-level knowledge graph and Bayesian theoretical reasoning.In order to better express the intricate relationship between the influencing factors,this thesis uses the knowledge graph to construct more comprehensive information,and uses the knowledge graph as a knowledge base to use machine learning to reason about the system state and locate the fault source.The constructed knowledge graph is used as data support for subsequent fault diagnosis.The richer the knowledge graph content,the more accurate the reasoning can be.The emphasis and difficulty of the scheme proposed in this article is how to construct a rich knowledge graph.(2)In order to enable the constructed knowledge graph to provide strong data support for fault diagnosis,this thesis proposes a method for constructing a multi-level knowledge graph.Analyze the levels that affect the state of the system and obtain multisource data according to each level.Construct a single-level knowledge graph based on data characteristics,and then use a multi-source data fusion model to fuse data at all levels to form a multi-level knowledge graph with comprehensive content coverage and threedimensional structure.(3)Due to the problem of missing relationships in the knowledge graph,inaccurate fault diagnosis occurs.This thesis proposes an ENCProj E model to mine its implicit knowledge and make the content coverage in the multi-level knowledge graph more comprehensive.The ENCProj E model integrates the rich semantic information contained in the multi-level knowledge graph into the Proj E model to enhance the model's link prediction ability and mine the knowledge of the multi-level knowledge graph.In order to solve the problems of fault diagnosis in complex industrial fields,this thesis proposes a complete fault diagnosis solution based on knowledge graph and machine learning.Multi-level knowledge graph is used as the data support for fault diagnosis,and the method based on the combination of knowledge and data drive is used for fault diagnosis,which provides a new idea for fault diagnosis.Therefore,the research in this thesis has a certain positive and practical significance.