Fault Prediction Research Of Turbofan Engine Based On Data-Driven Methods

Since the 21 st century,data-driven fault prediction technology has developed rapidly at home and abroad with unprecedented momentum.Soon it was highly regarded by the government,academia and industry.It has been widely used in many fields such as aerospace,machinery,electronics,ships,and automobiles.Aircraft engine failure is one of the major threats to aircraft safety flight accidents.As the most important type of aircraft engine,turbofan engine has a variety of fault mechanisms and fault models due to its complex working mechanism,frequent changing conditions and composite system structure.It is of great significance to predict the failure of turbofan engines to ensure flight safety and reduce maintenance costs.In view of this,this paper intends to study along the logic of the “Analysis on characteristics and fault of turbine engine-Fault feature extraction-Fault model recognition and fault diagnosis-Remaining Useful Life(RUL)prediction”.The main research contents are summarized as follows:The research background of the turbofan engine was analyzed,and both theoretical and practical significance of the paper was explained.The further summary and review on the existing research results at home and abroad were carried out from the three perspectives of remaining service life prediction,fault identification and turbofan engine failure prediction.The related characteristics and the main classification of Turbofan engine,coupled with its structure and the function of each component were specifically analyzed.Furthermore,the basic principles of fault mode and fault diagnosis were summarized,and the main parameters of fault occurrence were concluded.In order to improve the prediction efficiency and accuracy,Denoising Auto-Encode,KPCA and StandardScaler were used to conduct the reductions of noise and dimension of the original data samples,and nondimensionalization.CDT-RF model was constructed and applied to the prediction of fault pattern recognition and fault diagnosis in turbofan engines.Compared with the RF model,it improves the processing performance of continuous values and has higher accuracy and stability.The LSTM algorithm was optimized using GridSearch-NI and applied to predict the remaining service life of turbofan engines.Through the comparison experiment,it is concluded that the constructed model can avoid the gradient disappearance while grasping the characteristics of time series data.The residual service life prediction of turbofan engine hasbetter accuracy.