神经网络预测方法在鱼雷系统传感器故障诊断中的应用

Application of Neural Network Prediction to Sensor Fault Diagnosis for Torpedo Control System

介绍了基于神经网络的时间序列预测器的传感器故障诊断方法。首先采用神经网络建立预测模型,然后利用预测值和传感器实际输出值之差判断传感器是否发生故障。分别使用BP和Elman两类神经网络,经过离线学习和在线学习对某传感器偏移故障进行检测,仿真试验表明该方法对传感器故障诊断行之有效。

A method of detecting sensor failures through a neural network based time series predictor for single sensor was introduced. At first, a prediction model was built by a neural network. ; Secondly, the discrepancy between predictive value and practical output of sensor was compared to judge sensor failures exist or not. At the end of this paper,two kinds of neural network（BP and Elman） were used to build the on-line and off-line prediction models. Simulation result shows the method is effective for sensor fault diagnosis in torpedo control system.