径向基函数预测网络拓扑确定新方法  

Topology Determination Method for Radial Basis Function Prediction Network

作　　者：吕永乐[1] 

机构地区：[1]中国电子科技集团公司第14研究所,江苏南京210039

出　　处：《测试技术学报》2012年第2期171-178,共8页Journal of Test and Measurement Technology

基　　金：总装"十二五"预研基金资助项目(51317050202);国防"十一五"预研基金资助项目(102010201)

摘　　要：预测能力相对薄弱,已经成为制约PHM(Prognostics and Health Management)技术发展和应用的瓶颈.随着传感器和BIT(Built-in Test)设计技术的日益进步,采用序列分析的方法对复杂系统装备进行故障预测已经成为可能.在基于序列分析的预测方法研究中,径向基函数预测网络具有结构简单、学习速度快、具备非线性建模能力等诸多优点.为了改进其预测性能,在深入分析网络拓扑对模型性能及建模时间影响的基础上,综合考察了序列最佳线性自相关长度、建模精度和模型复杂度等多种因素,提出了基于偏自相关函数统计检测的输入层节点数目确定算法和基于BIC(Bayesian Information Criteria)准则的隐层节点数目确定算法,用以构建径向基函数预测网络;并对算法的有效性进行了分析.仿真结果表明,同传统建模算法相比较,由新算法构建的径向基函数预测网络具有最佳的预测性能,且建模时间不足传统算法的3%.With the progress of sensor and builtin test （BIT） technology, it is realizable to prognose the health status of complicated system by employing the methodology of time series analysis. In the research of prediction methods based on time series, the radial basis function prediction network （RBFPN） was widely paid attention because of the merits such as brief construction, fast learning and the nonlinear modeling ability. In order to improve the RBFPN prediction performance, the influence of network topology on the model performance and the consumed modeling time was analyzed. After that, the factors including optimal linear correlation length, the modeling precision and complexity were researched, and then the number decision algorithm of input layer nodes based on partial autocorrelation function statistical testing and the neuron number decision algorithm of hidden layer based on the Bayesian information criteria were respectively put forward to build RBFPN. Meanwhile, the validity of the algorithms was analyzed. Finally the results of simulation show that, compared to the RBFPNs with the traditional algorithms, the RBFPN constructed with the new algo- rithms has the best prediction performance and the modeling time is less than 3 % of the traditional consumed time.

关 键 词：故障预测 时间序列 偏自相关函数 径向基函数 预测网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] O213[自动化与计算机技术—计算机科学与技术]