Research On Intelligent Fault Diagnosis Method Of Gas Turbine Blade Based On Information Fusion

Gas turbine fault diagnosis technology has always been a hot spot in domestic and foreign research.Gas turbine status monitoring and fault diagnosis can greatly improve the safety and reliability of unit operation,and can also greatly reduce the maintenance cost,so the research on gas turbine fault diagnosis has important theoretical significance and application value.Aiming at typical blade fracture faults of gas turbines and based on test data of small gas turbine test bed,this paper carries out research on diagnosis methods and research and development of fault diagnosis expert system.Main research contents and achievements:(1)Feature extraction and feature compression methods of experimental test data.The fault test data of small gas turbine blade are analyzed,and the fault characteristics are extracted based on the analysis.The characteristic values are compressed by principal component analysis method,and the differences between the characteristic values of different faults are compared.(2)To study the three kinds of solve the small sample gas turbine based on information fusion method of fault data,including the BP neural network combined with DS theory of evidence method,support vector machine(SVM)method and the depth of the belief network method,the experimental data was used to extract the characteristic values of the input and output,as well as the effect of the three methods of diagnostic test validation;The characteristics of different methods are compared and analyzed.The results show that the fault diagnosis accuracy based on deep confidence neural network is the highest,reaching more than 90%.(3)Developing expert system for gas turbine shaft fault diagnosis.Based on the analysis of fault mechanism and vibration characteristics of gas turbine shafting,an expert database has been established and an expert system of gas turbine shafting faults has been developed and designed based on the analysis and statistics of corresponding characteristic values.