遗传神经网络在导航传感器故障诊断中的应用  

Application of Improved Genetic BP Network in Inertia Navigation Fault Diagnosis

作　　者：钱华明[1] 王雯升[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《中国航海》2009年第1期6-9,共4页Navigation of China

摘　　要：对船用组合导航系统中占主要地位的惯导系统进行故障诊断研究,提出了基于遗传算法优化的神经网络故障诊断模型。利用改进的遗传算法对BP神经网络的权值和阈值进行遗传操作,获得具有一定遍历性的初始权值和阈值,然后再利用神经网络的L-M训练方法进行训练,克服了BP神经网络搜索速度慢和容易陷入局部极值的缺点,保证了训练过程收敛。对采集到的传感器故障信号进行预处理,故障特征信号作为改进遗传神经网络的输入信号来检测故障。采用惯导系统陀螺仪故障信号进行仿真研究,结果表明此方法可有效地检测故障且提高了系统精度。The fault diagnosis of inertia navigation system which plays an important role in ship integrated navigation system is studied.The method chosen in the fault diagnosis is combined Genetic Algorithm and neural network.Ergodic initial value of weight and bias is searched for further training by introducing Advanced Genetic Algorithm,which improves search efficiency and global convergence of the network.Collected fault signal is preprocessed to produce fault characteristic frequency vector as the input of fault diagnosing network.The fault signal of gyro which is the crucial equipment in inertia navigation is taken as an example of simulation,of which the results indicate that this method can diagnose faults effectively and improve system precision.

关 键 词：船舶、舰船工程 遗传算法 神经网络 惯性导航 陀螺仪 故障诊断 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置] U666.12[自动化与计算机技术—控制科学与工程]