Research On Fault Diagnosis Of Aircraft Actuator Based On Data Driven Method

With the development of computer science,more and more digital devices are put into industrial production and applications.A large amount of historical data are generated during the operation of the device.These historical data will reflect the operation status of the equipment,and the abnormal data represents the equipment failure.The actuator is the active control component in aircraft flight control system.The actuator controls the elevator of the aircraft,ensuring the stable flight attitude of the aircraft.Once the actuator fails,the control of the aircraft’s elevator surface will be greatly difficult.The failure of actuator has a serious impact on the entire flight control system of the aircraft.Therefore,the study of real-time monitoring and fault diagnosis of aircraft actuators are of great significance.In this paper,based on the data-driven method,on the basis of fully understanding the physical structure of the actuator,the aircraft actuator is simulated in the Simulink environment,and a nonlinear model of the aircraft actuator in four different states is established and simulated.In this model,the signal source of each component of the aircraft actuator are included.Perform signal processing,feature extraction and fault diagnosis on multiple sets of data samples in sequence,the main work is as follows:(1)Obtained the initial signal in the Simulink model of the actuator,and used three wavelet threshold algorithms to reduce the noise of the initial signal.The simulation experiments are used to compare the signal-to-noise ratio and mean square error of the three methods,and the hard threshold algorithm is used.(2)The noise-containing signal was decomposed by EMD to extract the characteristics of the actuator signal in different states.IMF energy feature vectors are used to identify and diagnose actuator failures.IMF energy feature vectors of actuator signals in different states are significantly different.Qualitative analysis and preliminary classification of actuator faults are performed..(3)Based on the noise reduction and EMD decomposition of the actuator signal,the time-frequency diagram of the actuator signal is drawn,and the actuator failure is quantitatively analyzed and recognized based on the convolutional neural network.The VGG16 convolutional neural network model selected in this paper has an accuracy rate of 95.4% for actuator fault diagnosis after training.After model optimization,the recognition accuracy of the new VGG16 model has been increased to98.05%,and the model calculation time is extremely short compared with traditional methods.The data driven method of actuator fault diagnosis has been greatly improved in terms of accuracy and timeliness.In this thesis,a complete set of actuator fault diagnosis model is established based on the data-driven method,which realizes the real-time detection of aircraft actuator faults and has certain application value.