基于小波神经网络的水下航行器传感器故障诊断

Fault Diagnosis of UV's Sensors Based on Wavelet Neural Network

针对水下航行器系统的传感器故障诊断问题,提出了一种基于小波神经网络的传感器故障诊断方法.在对水下航行器系统的传感器故障信号进行特征提取时,发现其大部分能量都集中在低频部分.若直接以此能量分布来区分正常与故障、故障与故障的信号,将导致神经网络训练时间和分辨时间都会很长,因而不能实时地监控系统.为了很好地进行区分,凸显其差异,将低频部分能量舍去,只保留其余部分,并将其归一化,再利用径向基神经网络进行分类.利用小波分解的节点能量差异与特征提取特点以及神经网络的自我学习能力,通过大量的样本训练后,使神经网络很好地分辨出5类故障信号及正常信号.仿真结果表明:此方法简单、易于实现,适于水下航行器的传感器故障的诊断.

A fault diagnosis method based on wavelet neural network was proposed to diagnose the fault of sensors used in underwater vehicle （UV）. It was observed that the majority of the faulty signals＇ energy was concentrated in the lower frequency range. But, if this frequency distribution was taken directly as the basis on which to train the system and to distinguish the fault signals form the normal ones or to distinguish one kind of fault from another, it would take much long time of the system and real-time ability of the monitoring system would be ruined. A new method was proposed to carry on the classification with RBF neural network according to the rest part of signals which was normalized after the energy of lower frequency range was abandoned. With the nodes energy differences form wavelet decomposition algorithm, feature extraction characteristics the self-learning ability, after being trained with a large number of samples, the neural network could distinguish five kinds of fault and normal signals with high resolutions. The simulation results proved the simple and easy use of the method, which is suitable for the fault diagnosis of the sensors in UV systems.