摘要
针对测量中存在的陀螺随机漂移误差,提出了一种基于灰色RBF神经网络的预测建模方法.首先采用时间序列的饱和嵌入维数确定RBF神经网络模型输入层的节点数;其次采用灰色聚类法对输入样本进行分类,以确定RBF神经网络模型隐含层的初始节点数;最后采用灰色关联分析法对RBF神经网络的冗余隐含层节点实施删除,以得到满足精度要求的最小结构的RBF神经网络模型.将其应用到某型挠性陀螺随机漂移误差的预测建模中,可得预测模型的精度为90.33%,实验结果表明了该模型的有效性.
A combined method was applied to random drift model for a flexible gyro sensor to enhance its performance. The combined method is an improved RBF （radial basis function） neural network based on grey theory and analysis of time series. First the embed dimension of time series was used to select the number of input node for RBF neural network, and grey clustering method was used to select the initial number of hidden node for RBF neural network. Then grey correlation analysis method was applied to analyze the correlation degree between hidden node output and network output. According to the size of correlation degree, the redundant hidden nodes were deleted to realize the optimization of RBF neural network structure with a model accuracy of 90.33%. The experimental results show that the proposed model is capable of predicting.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2011年第2期39-42,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家985工程资助项目(000-X07204)
福建省自然科学基金资助项目(2010J05141)
