基于LMBP神经网络的故障预报方法及其应用  

A Fault Prediction Method Based on LMBP Neural Network and Its Application

作　　者：吕瑛洁[1] 胡昌华[1] 

机构地区：[1]第二炮兵工程学院,西安710025

出　　处：《机械科学与技术》2006年第1期28-30,共3页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家自然科学基金重点课题(69931040);国防预研课题(103010201)资助

摘　　要：利用具有二阶收敛效应的Levenberg-M arquardt(L-M)算法优化BP的权值修正量,提出了一种快速收敛的LMBP学习算法,给出了基于LMBP神经网络的故障预报方法,并应用于某陀螺仪的故障预报。结果表明:较之BP和传统改进BP网络,LMBP网络有着更高的收敛速度;基于该网络的时间序列预测模型可以实现性能优越的非线性预报器,将其应用于非线性系统的故障预报能够取得良好的效果。Through optimizing standard BP algorithm with Levenberg-Marquardt （L-M） has second-order convergence effects, the paper proposes a new learning algorithm algorithm which LMBP algorithm that converges very fast. A new fault prediction method based on LMBP neural network is presented, and its application to the fault prediction of a gyro is discussed. Experimental results show that the LMBP algorithm converges very rapidly compared with that of BP algorithm and improved BP algorithm. The LMBP neural network can predict nonlinear time series very well and the new method is effective for the fault prediction of nonlinear systems.

关 键 词：故障预报 LMBP神经网络 L-M算法 非线性时间序列预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]