Research Of Aircraft Unexpected Failure Diagnosis Based On T-S Fuzzy Neural Network

Aircraft unexpected failure is characterized with great uncertainties and ambiguity,traditional fault diagnosis methods can not deal with this question effectively. At present the domestic airlines commonly apply QAR(Quick Access Recorder)data to record the aircraft data that flying in the air. The detailed data of the true state of the aircraft in flight was recorded by QAR, these data provide a strong basis for the real-time status monitoring and fault diagnosis of the aircraft. Especially in the fast-growing aircraft maintenance workload,flight time is limited and troubleshooting, maintenance in condition of the lower efficiency, it is very important to conduct a deep research to improve the level of aircraft fault diagnosis.T-S fuzzy neural network technology is used in this paper. Combining with aircraft maintenance processes and taking QAR data as research object, we establish a fault diagnosis system based on T-S fuzzy neural network, which can locate and determine the severity of the fault and effectively help aviation maintenance personnel to quickly and accurately troubleshoot work. Mainly to complete the work in the following aspects:(1) A global analysis was studied by applying T-S fuzzy neural network to deal with fault diagnosis, the design goals has been determined, the probably problems may be encountered in the design has been forecasted; Finally,we design the overall framework of the system.(2) Genetic neural network algorithm was improved. For the case that when dealing with QAR data, BP neural network convergence showed poor even worse, we designed an error evaluation function to make a real-time dynamic monitoring for the training error values.When the error has little change or maintain a constant value, we introduced genetic algorithm to optimize the network weights and threshold. Giving full play to both the general convergence of BP neural networks, the change of genetic algorithms to optimize the network and make time to get an accurate guidance, it very greatly solved the problem of convergence of network diagnostics.(3) T-S fuzzy neural network diagnostic system network platform was builded. in the base of neural network, we constructed T-S fuzzy neural network by introducing fuzzy logic.A parallel sub-fuzzy network system was designed to solve fuzzy rules excessive proliferation issues ? ?caused by more input values; Through the output value of the fuzzy neural network we can locate fault location and determine the severity of the fault.In this paper, aiming to solve the aircraft unexpected failures diagnosis by dealing with QAR data, the modular division methods was used to simplify the treat principles, Theexperiment of aircraft unexpected failures diagnosis based on T-S fuzzy neural network was conducted. An improved genetic neural network algorithm was raised and double parallel network was designed to resolve the fuzzy rules disaster caused by too many imports.Experiments show that we got the desired objectives, the designed structure of the diagnostic system is capable of aircraft fault location and severity of the judgment.