Health Monitoring Method Of Aero-engine Based On Data-driven Technology

Civil aeroengine as a highly complex and precision machinery,is the "heart" of the aircraft.It works on the environment of high-temperature,high-speed,high-stress,strong-vibration and its working status and load changes constantly.Higher requirements are put forward on its reliability and security.More performance parameter containing the information of engine can be monitored with the technology development of sensor.However,the monitoring data and fault is nonlinear because of environmental and transmission noise,complex working environment.The gas path faults occupy more than 90% of aeroengine faults and its maintenance costs occupy 60% of the total maintenance costs.So,it's necessary to implement health monitoring to gas path components.Parameters reduction,parameters forecasting,fault diagnosis and fusion intelligent diagnosis were realized in this paper by analyzing the performance parameters of aeroengine.The main research contents are as follows:(1)The background of aeroengine health monitoring,the development history and status of technology,laws and regulations in national and international,the necessity and significance of carrying out health monitoring technology were concluded.(2)The preprocessing of gas parameters based on exponential smoothing and Savitzky-Golay smoothing is studied to eliminate the influence of noise data.Two smoothing results are compared(3)Aeroengine fault diagnosis method based on artificial immune algorithm is researched and a double population immune algorithm that can achieve threshold self-optimization is proposed.The aero-engine fault diagnosis is realized..(4)Aeroengine gas path parameter prediction based on ELM_Adaboost strong predictor was studied,and several indexes were used to evaluate the prediction results.(5)Based on the two-population immune algorithm and ELM_Adaboost strong predictor fusion prediction of aeroengine residual life,with five models of engine airway samples as an example,to achieve 2000 flight cycle failure state prediction.