Fault Diagnosis Of Aero-engine Based On Data-driven Technology

Predictive maintenance system is designed to improve reliability and safety of aero-engines. The essential and most important part of predictive maintenance system is the fault diagnosis system. The faults of aero-engines not only affect the safety and reliability of airplanes but also add great additional costs to national economy. Gas path faults are the most common form of aero-engine faults. It’s important to develop fault diagnosis technology of aero-engines, which can assess aero-engine conditions in real time and predict the possibility of faults in the future.This dissertation studies the fault diagnosis system of aero-engines based on data driven technology.Firstly, this dissertation studies the characteristics and performance of relevance vector machine and then uses the technology for fault diagnosis of aero-engine. The data used in the study is obtained from a mixed exhaust turbofan engine model. The diagnosis results are compared with those of the support vector machine.Secondly, the characteristics and performance of multiclass relevance vector machine is studied(M-RVM). The technology is used to establish a fault diagnosis method of aero-engine and the diagnosis results are compared with those of the relevance vector machine. The data used in the study is obtained from a mixed exhaust turbofan engine model.Finally, a hybrid fault diagnosis method using relevance vector machine and RBF artificial neural network is established. The relevance vector machine is used to determine which component is faulty and RBF artificial neural network is used to decide how serious the fault is. The data used in the study is obtained from a mixed exhaust turbofan engine model.