基于RBF网络的激光陀螺标度因数数据建模研究  

Data modeling of laser gyro scale factor based on RBF network

作　　者：邵玉梅[1] 赵忠[1] 王璐[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《中国惯性技术学报》2006年第6期63-65,共3页Journal of Chinese Inertial Technology

摘　　要：前向神经网络中的径向基函数(RBF)网络是一种局部逼近网络,它用局部逼近的总和达到对训练数据的全局逼近,在理论上可以实现全局最优。该文利用径向基函数神经网络对某一温度段的陀螺标度因数的温度数据进行建模处理,并利用各组数据建立一种两因素RBF网络,这两个输入因素选择为温度以及各个温度值对于所属组初始温度的增量。仿真结果表明,所建立的两因素RBF网络可以精确地拟合各温度下的标度因数温度数据,仿真数据的误差与均方差比用BP网络训练的数据效果要好,在数值上提高了近一个数量级。RBF(Radial Basis Function) network in foreword neural network is a local-approached network which uses local-approached summation to achieve global-approach.Based on the RBF artificial neural network,the paper set up a network model of temperature data for gyro’s scale factor.The network had two factors,one was the measured temperature and the other was the temperature-increment corresponding to the initial temperature of each group.The simulation results indicated that the established RBF network with two factors could accurately match with the scale factor-temperature data,and the data errors and mean squared errors were one order lower than that of BP(Back Propagation) network.

关 键 词：激光陀螺 标度因数 RBF网络 温度模型 

分 类 号：U666.1[交通运输工程—船舶及航道工程]