Modeling,simulation And Prediction Of Several Typical Faults Of Aeroengine

Aeroengine is a kind of aerothermodynamic system with extremely complex structure.As the "heart" of an aircraft,it provides the aircraft with a source of power for flight.Because of its complex structure and the susceptibility to environmental factors,aeroengine can trigger one or more failures in the long flight process,affecting the stability and safety of aircraft operation.Therefore,it is of great significance for improving the security of aircraft flight to study and analyze the modeling,simulation and prediction of typical faults that often occur during the operation of aeroengine.This paper mainly studies three typical faults of aeroengine,which are compressor surge,combustion chamber atomization and tail nozzle thrust drop.Due to the particularity of aeroengine itself and the confidentiality of key parameter data of typical faults in operation,therefore the research content is separated into three parts mainly:In the first part,the definition,causes,symptoms and hazards of three typical faults of engine are analyzed and summarized.In the second part,engine modeling and model-based simulation of typical faults.Firstly,the engine's inlet,fan,compressor,combustion chamber,high and low pressure turbine,the inlet and outlet of the mixing chamber and the tail nozzle are modeled and simulated by using the thermodynamic change mechanism during the operation.Secondly,using the power and flow balance equations that the components work together to build the whole engine model and simulate the normal operation process.Finally,three typical faults are simulated based on the component level model.According to the typical symptoms of three typical faults,anomalies or disturbances are added to the model input to simulate the changes of key parameters representing the typical faults.In the third part,the trend prediction of typical aeroengine fault data based on Simulation.Firstly,three prediction algorithms,SVM network,BP neural network and LSTM recurrent neural network,are constructed based on simulation data.Secondly,the evaluation index of fault prediction effect is put forward.Finally,the three neural network prediction algorithms are used to compare the surge fault,the combustion chamber atomization fault and the tail nozzle thrust drop fault for simulation prediction comparison,result analysis and performance evaluation.In this paper,a component level simulation model of aeroengine is built.Based on this model,the variation process of three typical key parameters of engine fault is simulated,and the prediction accuracy of the variation trend of key parameters of fault is compared and evaluated with the constructed neural network algorithm.This study provides a visual tool and means for the establishment of complex system model,also provides data support for aeroengine fault prediction,and provides a solution to the problem of difficult data acquisition for complex system.At the same time,it provides practical technical analysis methods and tool ideas for fault data trend prediction.This research work was supported by the National Natural Science Foundation of China(No.61973094,91646108).