Research On Fault Diagnosis Of PW4000 Based On SVM Multi-classification

It is very important for aero-engine using status monitoring and fault diagnosis to ensure its normal operation. Foreign countries have developed a variety of aviation engine diagnostic monitoring system that can provide reliable real-time monitoring and accurate diagnostic analysis for aero-engine's running by using monitoring technology and computer technology currenttly.The smoothing method of aero-engine's performance parameters is studied firstly. The fusion of multi-index evaluation method is proposed which can select the optimal smoothing method according to the actual needs of different performance monitoring parameters, having good practicability and generalization. Then the method of acquiring baseline model is studied for the baseline surveillance which use data from the monitoring system and the QAR based on SVR and the model can meet the accuracy requirements of baseline monitoring through the actual test analysis. Then the diagnosis of multiple faults based on multi-class SVM is studied and fault data of fingerprints figure is used and SVM multi-classification fault diagnosis model is built by ratio coefficient and correlation coefficient which shows that the classification model has better reliability. Finally, the aero-engine fault diagnosis system is established which use monitor and diagnostic techniques, including the extreme monitoring,differences monitoring, baseline monitoring, fault diagnosis of multi-class and other functional modules, providing reliable monitoring and diagnostic analysis for the aero-engine.The performance parameters smoothing evaluation method, the baseline model method and the multi-class fault diagnosis method are proposed by the use of data smoothing principle, SVR and SVM centering on aero-engine's diagnostic monitoring, providing accurate monitoring and reliable diagnostic analysis to detect unusual trends in time.