Navigation Sensors Fault Diagnosis Based On Wavelet Transform And Neural Network

Under the demand of aviation and space technology, A carrier may be equipped many different navigation equipments. And only a navigation equipment can not adept to all the circumstances, The information provided by Navigation system had developed from simple position parameter to the combination between navigation system and measurement system, which can provide accurate position and military measurement composed into a navigation system. Navigaiton system can compensate navigation information which provided from navigation equipments, with this ability intergrated navigation system can adept to complexity work environment and has accurate navigation accuracy.In the paper the integrated navigation sensors system is the subject investigated. The paper chose an intelligent method to fault diagnosis based on navigation sensors. When a sensor had faults other sensors instead the fault sensor as the navigation equipment, the method can avoid the high price of hardware redundant. Many materials are looked up before the theme start, and proof the theme feasible. In the paper analyzed the process of fault diagnosis and the purpose of fault diagnosis, and introduced wavelet transform.The paper simulates the wavelet transform and neural network process, which target is gyro owing to inertia system. Sampling the status output of gyro, decompose the signal by three-layer wavelet packet. Feature exertion after decomposition for eight nodes, then an 8-dimensional eigenvector is used as fault samples to train a three-layer Radical Basis Function (RBF) neural network. After the training process the network can detect a fault on-line. After testing the method can detected the faults accurately.Compactness method was chosen to fault diagnosis, the method mixes the wavelet transform and neural network together, which choose wavelet function as the neuron function of BP neural network. The GPS and DVL were chosen as the examples to faults diagnosis. The method avoids the disadvantage of BP neural network which is local minimization problem. The efficiency of wavelet neural network enhanced 50% compared with the BP neural network. After testing the method can detected faults diagnosis exactly.