Aero-engine Fault Diagnosis And Prediction Under Complicated Operating Conditions

There are many parts of aero-engine,and the complicated working conditions make aero-engine more prone to failure.As the operating time increases,the performance of aero-engine will decline,and its reliability will continue to decline.If potential engine failures cannot be found in time,major disasters such as aviation accidents may be caused.Fault diagnosis and performance prediction of aero-engine is an effective means to reduce catastrophic accidents caused by faults and reduce maintenance costs,and it is also one of the important research topics in the aviation field.Based on a data-driven method,this paper studies the fault mode diagnosis and performance parameter prediction methods of aero-engine under complex operating conditions,as follows:(1)Aiming at the problem that aero-engine monitoring data changes under multiple operating conditions is more complex and traditional machine learning algorithms need to use feature fusion technology to improve diagnosis accuracy,this paper proposes an aero-engine fault mode diagnosis method(EFMC),which uses a one-dimensional convolutional neural network model(1D-CNN).First of all,the engine monitoring data is standardized according to the type of working condition parameters.Then,through the convolutional layer and pooling layer in the 1D-CNN,the degradation features in the original sensor monitoring data are extracted,and the fully connected layer integrates all the degradation features to obtain the fault mode classification results.Finally,this paper verifies that the EFMC method has high diagnostic accuracy by comparing with support vector machines,bidirectional long short-term memory networks,and long short-term memory networks(LSTM)combined with attention mechanism models.(2)Aiming at the problem that the key performance parameters of the aero-engine will affect its decay and the complete data is not easy to obtain,this paper proposes an aero-engine performance parameter prediction method(SEMP),which uses the bayesian ridge regression model.Before predicting the parameters,the local outlier factor algorithm(LOF)is used to eliminate outliers,which further improves the prediction accuracy.Finally,this paper demonstrates that the SEMP method has high prediction accuracy by comparing the prediction accuracy with ridge regression,K-nearest neighbor,extreme learning machine,LSTM and gated recurrent unit network.This paper uses the C-MAPSS dataset to perform experimental verification and comparative analysis of the proposed methods.The experimental results show that the diagnostic accuracy rate of the EFMC method reaches 99%,and the coefficient of determination R~2 reaches 0.99 after using the SEMP method,which is better than other comparative models,and has certain engineering development and theoretical reference value for improving the operational reliability of aero-engine.