Research On Aircraft Subsystem Fault Diagnosis Based On Fuzzy Neural Network

With the rapid development of computer neural network technology and the backward status of fault diagnosis in modern combat aircraft, fault diagnosis of automation and intelligent is calling for the gradual increase. How well complete the fault diagnosis in aircraft equipment based on fuzzy neural network has gradually become an important research direction.A certain type of aircraft subsystem is a nonlinear system. Nonlinear system identification based on fuzzy neural network is achieved through structural identification and parameter identification.Identification of the structure involves the input and output of the space division, the number of fuzzy rules and the premise and conclusion of the initial set of parameters. Combined the limitations of the fuzzy c-means clustering algorithm for accessing the fuzzy rules with the advantages of particle swarm optimization, the improved algorithm is rasied which is based on fuzzy c-means clustering algorithm. This algorithm can be used to anti-noise, anti-sensitive and achieve global optimization, improved efficiency and the clustering effect. According to the clustering results, the fuzzy rules is developed. Then the initial structure of fuzzy neural network is generated.The four-match fuzzy neural network matching with a fuzzy inference mechanism is constructed in the parameter identification. According to the BP learning algorithm, the parameter value in the initial network Structure is constantly adjusted. The process generates a higher precision fuzzy neural network. With the diagnosis and the actual situation, the fault diagnosis model of aircraft subsystem is designed. The aim is achieving better results in the fault diagnosis.In the implementation of the system, the system overall framework is designed. And the functional modules are realized which includes the sample pretreatment module, the feature extraction module, the learning training module and the reasoning diagnosis module. As testing and analysising the improved algorithms and the fault diagnosis model , their validity and accuracy are verified.