Performance Parameters And Remaining Useful Life Prediction Of Aeroengine Based On Deep Learning

The health of aeroengine is directly related to the safe operation of aircraft.Due to the complex structure and variable operating environment of aeroengine,physical models of health monitoring are difficult to establish accurately.Therefore,it is of great significance to study artificial intelligence performance prediction methods using engine operating data.In this paper,the data-driven method is adopted to study the prediction of engine performance parameters and remaining life.The main research contents and innovative achievements are as follows:Firstly,a deep convolutional neural network model for exhaust temperature prediction is proposed.First of all,the aeroengine operating data provided by an airline company is preprocessed,and the time series sample set is constructed by the method of sliding window according to the length and feature number of the data.Then,the specific structure of DCNN prediction model is designed considering the number of pooling layers and convolution layers,and the length of time window.Finally,the data of the CFM56-7A engine is used to verify the prediction effect of the proposed model.Experimental result shows that the proposed model can automatically extract the deep features hidden in the original data and predict the exhaust temperature of the aeroengine.Secondly,a DCNN-LightGBM model for exhaust temperature prediction is proposed.First of all,the principle and structure of Light Gradient Boosting Machine make it have excellent decision effect.Then,using the advantages of LightGBM,the DCNN-LightGBM combined prediction model is established based on the forementioned DCNN model.The combined model can reduce the probability of falling into the local optimal value and improve the prediction ability.Finally,the result shows that the combined model has higher prediction accuracy and stronger robustness than other methods.Thirdly,the DCNN-LightGBM model for remaining useful life prediction is proposed.First of all,the original engine data is preprocessed and the features is selected.Then,by designing the model structure and selecting the loss function,the DCNN-LightGBM model is established to approximate the real remaining life of the engine.Finally,the result shows that the model can provide more accurate remaining life of the engine compared with other classical methods.In summary,to solve the forecasting problem of exhaust temperature and remaining life for aeroengine,the DCNN and DCNN-LightGBM methods are proposed in this paper.These methods can predict the engine performance to ensure the safe operation and reduce the maintenance cost of aircraft.