船用计程仪信号传输系统的神经网络故障诊断算法  

A fault diagnosis algorithm for the signal transmission system of ship log based on neural networks

作　　者：徐国栋[1] 王俊雄[1] 崔国友[2] 

机构地区：[1]上海交通大学动力装置及自动化研究所,上海200030 [2]海军902厂,上海200083

出　　处：《舰船科学技术》2013年第8期115-118,共4页Ship Science and Technology

摘　　要：计程仪信号传输系统是由步进电机通过齿轮带动多个旋转变压器。为实现对其故障的智能诊断,提出一种基于BP神经网络算法的故障诊断方法,并采用LM(Levenberg-Marquardt)最优化算法对BP网络进行训练。通过Matlab神经网络工具箱仿真计算,结果表明LM算法提高了BP神经网络的学习效率及稳定性,明显加快了网络的收敛速度。使用BP神经网络对步进电机带动多旋变系统故障诊断的方法行之有效,在减少故障诊断人员工作量同时提高了系统故障诊断的效率。The signal transmission system of ship log contains many resolvers driven by a stepper motor through gears. In order to achieve its intelligent fault diagnosis, this paper proposed a fault diagnosis algorithm based on BP neural networks and trained it with LM （Levenberg-Marquardt）algorithm. The simulation through Matlab neural network toolbox showed that this method which was effective and accurate improved the efficiency and stability of the BP neural network and sped up the convergence rate of the network. It could reduce the workload and improve the efficiency of the fault diagnosis.

关 键 词：BP神经网络 旋转变压器 故障诊断 

分 类 号：U666.152[交通运输工程—船舶及航道工程]