Research On Key Technologies Of Turbine Intelligent Diagnosis And Health Management

With the introduction of national strategies such as "made in China 2025","Internet Plus" and "new generation of artificial intelligence development plan",smart power plants have become the future trend of energy enterprises driven by artificial intelligence technology.Intelligent Diagnosis and Health Management(IDHM)technology requires the integration of a new generation of artificial intelligence technology to help improve the efficiency of equipment operation,Diagnosis and maintenance in power plants.At present,there are many researches on fault diagnosis and maintenance decision,and each has its own characteristics.On the other hand,there is a lack of research on constructing IDHM technology system for complex system.As a result,various new technologies and methods cannot be effectively understood and applied by power plants.Especially for such important equipment as steam turbine,it is necessary to use IDHM technology to solve traditional problems and flexibly use experience knowledge to make IDHM technology more intelligent,fully considering the characteristics of complex turbine fault mode,little monitoring information,scarce fault samples,and abundant diagnostic knowledge.Therefore,in this paper,based on the original intention of IDHM research on realizing abnormal detection,fault diagnosis and equipment risk reduction by auxiliary personnel,on the basis of summarizing the difficult problems in turbine fault diagnosis,IDHM key technologies of comprehensive utilization of empirical knowledge and machine learning were studied.Firstly,based on the characteristics of strong empirical knowledge dependence,high knowledge re-usability and unstructured knowledge data in fault diagnosis and maintenance of steam turbine units,the analysis methods and steps of fault mechanism are summarized based on Equipment Tree Analysis,Fault Mode and Effect Analysis and Fault Tree Analysis.Based on the Knowledge Graph and Ontology theory,the construction process of fault diagnosis knowledge graph with complex relationship structure is proposed.Taking the nuclear turbine as an example,the fault diagnosis knowledge graph is established.By using knowledge graph to store and express diagnostic knowledge,the redundancy of knowledge data in the system is reduced,and the management efficiency of knowledge data in IDHM system is improvedSecondly,on the basis of summarizing the fault characteristics of the state data commonly used in fault diagnosis of steam turbine,the identification methods of trend symptom and spectrum symptom were studied.This paper presents a method of trend feature quantification of sequential data combined with experience,which makes up for the shortcomings of the previous method of identifying trend symptom in turbine fault diagnosis.Based on the principle that the excitation force propagates in the nonlinear system when the vibration fault occurs,a spectrum identification method based on the direction of the source is proposed in this paper.Based on the knowledge of steam turbine fault mechanism knowledge,the symptom identification method research in this paper makes up the shortcomings for identification method of trend symptom and spectrum symptom in steam turbine fault diagnosis,which is helpful for IDHM system to realize automatic symptom identification and improve system diagnosis efficiency.Thirdly,in order to make up for the defect of false alarm rate and missing alarm rate in symptom recognition method,and the failure of machine learning method to carry out knowledge reasoning and get fault causes and repair Suggestions,fault isolation,fault diagnosis and fault severity assessment methods are studied in this paper.n this paper,a fault isolation method based on graph database search technique is proposed to solve the problem of over-large diagnostic target range caused by redundant measurement points,excessive symptom information and reuse of diagnostic knowledge.Through fault isolation,the target range of subsequent fault diagnosis is greatly narrowed.n order to further deduce the possibility of fault occurrence,based on the concept of fault causal network,the fault diagnosis knowledge in the knowledge graph is transformed into a Bayesian Network diagnosis model for fuzzy reasoning.The interactive reasoning between the diagnosis system and the maintenance personnel is realized based on the online symptom and the manual troubleshooting information.In order to comprehensively evaluate the current operation risk level of equipment and optimize the fault troubleshooting sequence,a method for calculating the severity of fault chain was proposed,which comprehensively evaluated the possible fault chain in the diagnosis network from multiple angles,so that the maintenance suggestions could quickly reduce the operation risk level of equipment under fewer maintenance times.Finally,based on the above research,through the development of IDHM prototype system for nuclear turbine,the architecture,data warehouse and the main functions of IDHM system are designed and developed,so that the functions and data flow of various technologies can be effectively integrated.Through the development and test of the prototype system,the feasibility and effectiveness of the research content of this paper are verified.