Engine Fault Diagnosis And Prediction Based On Neural Network

With the rapid development of civil aviation industry in recent ten years,the safety problems of aircraft itself and in flight,which have been paid close attention to,are particularly important.As the core component of aeroengine,the safety and reliability of rotating shaft rotor directly determines the safety of aeroengine and aircraft.Therefore,it is necessary to study the fault of engine bearing.Firstly,in this paper,six kinds of typical bearing faults are set up based on the classic fault types of engine.In the fault experiment,the corresponding fault data are collected,and the simulation experiment and data collection of bearing fault are completed.Secondly,using the statistical signal processing method of signal processing technology,seven kinds of fault eigenvalues of the collected fault data are extracted.A new fault discrimination factor is proposed.Based on eight kinds of fault feature data,the fault data set is compressed,and the data volume is effectively compressed.On this basis,BP neural network is used to complete the diagnosis of six kinds of faults,and the best hidden layer node number and the best diagnosis effect of BP neural network,BP softmax learner,are obtained by cross validation and the method of forming a learner.However,it is still found that two of the six types of faults are misdiagnosed.After factor analysis,the matrix factor distribution of the two rotation components is the same.The reason of misdiagnosis was explained.The results of factor analysis also prove that the new discriminant factors have the value of characteristic contribution.Finally,this paper realizes the bearing fault prediction based on Elman neural network,verifies the memory characteristics of Elman neural network,and obtains the value that Elman neural network can be further applied in engineering design.