Research On Application Of Artificial Intelligence Method In Aero Engine Fault Diagnosis

As a long-term work in the harsh environment of high temperature,high pressure,strong vibration and large stress,the aero engine as the heart of the aircraft belongs to the typical fault-prone system.How to efficiently and accurately troubleshoot aero-engines has become one of the research topics that attract much attention in the industry.Therefore,this thesis takes a certain aero-engine air circuit system as the research object and carries out the research on fault diagnosis methods.The main contents are as follows:(1)Research on aero-engine parameter reduction method based on neighborhood rough set.This method combines the parameters of the engine gas system monitoring parameters to reduce the parameters of the gas circuit to obtain the characterization parameters that can represent the state information of the engine,and provides a basis for parameter selection for subsequent research on fault diagnosis methods.(2)Research on KPCA-WNN-based aero-engine fault diagnosis method.In this method,KPCA is used for feature extraction of the characterization parameter data,and the extracted data is used to establish a WNN fault diagnosis model.The results show that this method can effectively diagnose the faults of the engine air system.(3)Research on DAEN-based aero-engine fault diagnosis method.This method directly uses the characterization parameter data to establish a DAEN fault diagnosis model.The results show that,compared with the KPCA-WNN diagnosis method,this method can more effectively diagnose the faults of the engine gas system and has better diagnostic performance.(4)Research on aero-engine fault diagnosis method based on D-S evidence theory.This method uses characterization parameter data to establish two KPCA-WNN and DAEN fault diagnosis models,and uses D-S evidence theory to make decision-level fusion of the diagnosis results of both models.The results show that the D-S evidence theory can effectively fuse the diagnosis results of the two fault diagnosis models and achieve the purpose of improving the diagnosis effect.(5)Three common aero-engine fault diagnosis methods are used for fault diagnosis to verify the diagnostic performance of the fault diagnosis model studied in this paper.Using the same data,three fault diagnosis models are established and tested.The results show that the diagnosis performance of the fault diagnosis model studied in this paper is superior to traditional SVM,RBF and GRNN fault diagnosis models.