The Research Of FOG Testing And Intelligent Fault Diagnosis System

With the development of the civil aviation industry, as the core of the fiber optic gyro inertial navigation system, the requirement of Civil Aviation airborne FOG performance has become more and more high. To ensure the aircraft flight safety, the research for FOG testing and fault diagnosis is particularly important. And some intelligent fault diagnosis methods give us new ideas to solve the problem. For the study of the fiber of optic gyroscope, the platform of the intelligent fault diagnosis system is three-axis turntable based on wavelet technology, particle swarm algorithms and neural networks. It needs to analysis FOG works and error indicators, and give the FOG failure mode. It includes mainly analysis of the basic principles and the wavelet transform, the collected signal characteristics. To achieve the better signal analysis, firstly using the collected FOG signal wavelet to remove noise, and then the next step is using higher frequency resolution signal for decomposition of wavelet packet further. The wavelet packet decomposition of fiber-optic gyro output signal under various failure modes and energy rate distribution do not understand bands eigenvectors. To train the network has the ability to map the relationship between the fault and the characteristics,diagnosing network input needs to be extracted. Comparing performance and convergence speed between extensive BP neural network and RBF neural network, the latter is used to achieve fault diagnosis network. The improved particle swarm algorithm is to optimize the network to adjust and overcome the shortcomings of the RBF neural network, and the network performance indicators can show the different network performance merits. After the last run of the test system, to train and test the effect of this improved neural network,using a wavelet transform to obtain the training data of this improved neural network after optimization and accuracy.