基于分形和支持向量机的装备技术状态预测模型  

Forecasting Model of Equipment Technique Condition Based on Fractal and Support Vector Regression

作　　者：郭磊[1] 郭金茂[1] 徐达[1] 武新星[1] 

机构地区：[1]装甲兵工程学院兵器工程系,北京100072

出　　处：《科学技术与工程》2009年第17期5172-5175,共4页Science Technology and Engineering

摘　　要：基于分形和支持向量机回归理论,建立了装备技术状态预测模型。将反映装备运行状态的特征数据作为时间序列,首先进行相空间重构,得到时间序列的最小嵌入维数,以此作为支持向量机输入节点数。利用支持向量机对样本训练,建立预测模型。以装备振动信号预测为实例,表明将时间序列最小嵌入维数作为支持向量机输入节点数目,所建立的模型是最优的。支持向量机预测结果和真实值相比误差较小,可以满足装备技术状态分析和预测的要求。A technique condition forecasting model of military equipment based on fractal theory and support vector regression （SVR） was presented. Take the feature data that reflect equipment operation condition as a time series, techniques of phase space reconstruction were used to calculate the minimum embedding dimension, which regard it as the input nodes of support vector regression. The forecasting model was built up after samples series training. The case study of equipment vibration signal forecasting shows that, take the time series minimum embedding dimension as the input nodes of support vector regression, the model is optimal. Compared with true value, the SVR forecasting result is estimated with less error, the model can better meet the requirement of equipment technique condition analysis and forecasting.

关 键 词：技术状态预测 分形 相空间重构 支持向量机回归 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]