The Sensor Fault Diagnosis of the UV Based on the Wavelet Neural Network

To aim at the fault diagnosis problem of the underwater vehicle's (UV) sensor, an approach of sensor's fault diagnosis based on the wavelet neural network is proposed in this paper. When the UV system sensor's faulty signal is extracted, a majority of energy (over 90%) detected concentrates in the low frequency part. If directly discrimination between normal and fault, fault and fault by this energy distribution, the train of the neural network and discrimination will costs very long time, so the system can't be real-timely monitored. For nicely carrying on a distinction, highlighting its difference, the low frequency parts of energy must be remove, and the rest part will be reserved, normalized and classified with the RBF neural network. Then by using of difference of the node energy in wavelet analysis, abstraction of characteristic and self-learning ability of the neural network, the neural network has higher resolution to five kind of faulty signals and normal signal after a great deal of sample train. The result proves that this method is simple, easily implemented and suitable for classification of sensor's fault in the UV system.