Research And Implementation Of Fault Diagnosis Method For Flight Control System

Flight control system is the control core of aircraft,which plays a vital role in the implementation of aircraft missions.Once failure occurs without timely detection and correct diagnosis,the value of aircraft and the safety of personnel onboard will be fatally hit.Therefore,it is extremely important to conduct timely detection and correct diagnosis for flight control system.With the increasingly development of technology in the aviation field,while taking more and more important tasks,the flight control system has a more complex structure and a larger internal system.The fault diagnosis technology for flight control system is put forward higher request,therefore,it is of great significance to study more effective fault diagnosis method to adapt to the rapidly developing flight control system.Based on this background,the fault diagnosis mechanism of flight control system with an aircraft as the object was analyzed and studied in this thesis.The RBF neural network algorithm to improve the fault diagnosis accuracy of flight control system was focused on studying and improving.Then,the fault diagnosis software of flight control system was designed and implemented to assist maintainers in daily troubleshooting.The main research contents and work of this thesis are as follows:Firstly,the working principle,structure composition,redundancy configuration and fault detection mechanism of flight control system with an aircraft as the object are studied.Based on the characteristics of complex structure and huge system of flight control system,the source,representation and corresponding relationship and nature of the fault of flight control system are analyzed.Also,the way to acquire fault information and the first method of fault diagnosis are studied.Then,the fault diagnosis method for flight control system is further studied.For the fault of complex flight control system module,the reasons for using neural network method for diagnosis and classification are analyzed.In addition,the BP neural network and the RBF neural network commonly used in diagnosis and classification are studied and compared,and RBF neural network is selected as the model for fault training and diagnosis and classification of flight control system.Several training algorithms of RBF neural network are compared on the fault diagnosis and classification of flight control system,and the k-means clustering method is determined to study further to improve the ability of diagnosis and classification.The optimization idea of K-means clustering method and the optimization choice based on the direction of data density are analyzed and improved,then,the algorithm of this thesis is formed.The RBF neural network based on this algorithm is used to compare the training and diagnosis of the public dataset and the flight control system fault dataset.It is verified that this algorithm can improve the fault diagnosis and classification ability for flight control system to some extent.Finally,according to the actual requirements,the software of fault diagnosis for flight control system is designed and implemented,which includes background fault information acquisition module,fault information database module,graphical fault tree module and fault diagnosis module.This software has the ability of establishing graphical fault tree,and using graphical fault tree and neural network to diagnose and classify the fault of flight control system.At present,this software has been successfully applied to the routine troubleshooting for flight control system of an aircraft.