Research On Fault Diagnosis Of Aircraft Electromechanical System Based On Multi-Classifier Fusion

Aircraft electromechanical system is an important part of aircraft.It plays a very important role in the safety performance and the smooth flight of the aircraft.Electromechanical system failure is one of the main causes of aircraft accidents.The amount of sensors data is huge and complex and it is difficult to achieve ideal diagnostic results using fault diagnosis schemes based on a single classifier model and the diagnosis accuracy is difficult to be further improved,and the false alarm rate and missed judgment rate are relatively high.Based on the full analysis of the common fault signals of aircraft electromechanical systems,This thesis delved into the data fusion technology in fault diagnosis,studies the aircraft electromechanical system fault diagnosis scheme based on multi-classifier fusion.The main tasks are as follows:Firstly,the using of Dempster-Shafer(DS)evidence theory is researched and the shortcomings of traditional DS evidence theory are analyzed.In view of the poor fusion effect of the classic DS theory in the case of "Zadeh paradox","one-vote veto",and poor integration in situations of high conflict of evidence,This thesis proposes an improved DS evidence theory algorithm based on cosine similarity,using cosine similarity as the basis for assigning evidence weights,which significantly reduces the conflict between evidence and improves the fusion effectSecondly,aiming at the problem of difficulty in improving the diagnostic accuracy of the single classifier model,and the high false alarm rate and missed judgment rate,a fault diagnosis scheme for aircraft electromechanical system based on the fusion of multiple classifiers is proposed.According to the Q statistics,convolutional neural network(CNN),recurrent neural network(RNN),and long short-term memory network(LSTM)are selected as sub-classifiers.Then the results at the decision-level are merged based on DS improved by cosine similarity to completes the fault diagnosis.The false alarms and missed judgments of the system are significantly reduced,and it has a significant effect on improving the accuracy of fault diagnosis.Thirdly,the proposed fault diagnosis method is engineered to realize the fault diagnosis software of the aircraft electromechanical system.The software is divided into data processing module,CNN fault diagnosis module,RNN fault diagnosis module,LSTM fault diagnosis module,DS evidence fusion module.By comparing the software’s accuracy,false alarm rate,and false alarm rate with the previous test results,it is proved that the software has a better fault diagnosis effect.In summary,the process from theoretical research to software implementation and verification in this paper proves the effectiveness and superiority of the proposed fault diagnosis method of aircraft electromechanical system,which can effectively improve the work efficiency of ground maintenance personnel.