Research On Thchnology Of Fault Diagnosis For Fly-By-Light Control System Based On Neural Network

The Fly-by-Light (FBL) system is proposed for the shortage of the traditional Fly-by-Wire controlsystem. The high reliability of the flight control system is important to ensure flight safety. We studythe thchnology of fault diagnosis for Fly-by-Light control system, when the aircraft has faults oraccident injures, it is great important that the control system can quickly diagnose and isolate the faultto ensure the continued implementation of flight mission or the safe return.Firstly, the relevant theoretical knowledge of Fly-by-Light control system and fault diagnosis isdescribed in this thesis, and then we establish a model of three redundancy Fly-by-Light controlsystem and fault generation module based on the structure and common faults of the system.Secondly，the paper proposed a fault diagnosis method based on BP(Back-Propagation) network onthe basis of the model we have established. Then the simulation examples show the method has acertain effect.However, due to the structural characteristics of the BP network,when training sample islarger and the required accuracy is high, the network often do not converge and easy to fall into localoptimal solution.At last,for the lack of BP network, the paper also proposed a fault diagnosis method based onprobabilistic neural network（PNN）. The classification problems are resolved by using linear leamingalgorithm to replace the former work done by the nonlinear way.The high accuracy of nonlinearalgorithm is maintained at the same time. And the corresponding weights of the network are thedistribution of the samples.The network does not need to be trained，which can meet the real-timeprocessing requirements. Then the simulation examples show that this method based on theprobability neural network is better than BP network system in the speed，accuracy and the ability forsamples superaddition.