基于改进SVM的坑道毁伤仿真训练样本约简模型  

Training sample reduction model of damage simulation of protective engineering based on improved SVM

作　　者：袁辉[1] 王凤山[2] 许继恒[1] 沈英[3] 

机构地区：[1]解放军理工大学国防工程学院,江苏南京210007 [2]解放军理工大学野战工程学院,江苏南京210007 [3]中国电子科技集团第十四研究所,江苏南京210039

出　　处：《解放军理工大学学报（自然科学版）》2014年第2期152-157,共6页Journal of PLA University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(51308541);江苏省自然科学基金资助项目(BK20130066);中国博士后科学基金资助项目(20110491844)

摘　　要：针对坑道工程动荷段毁伤仿真科学实验训练样本中存在的相似样本和大量对分类器模型构造"无用"的冗余信息,提出了一种基于改进支持向量机的动荷段毁伤仿真实验训练样本约简方法。围绕约简任务和功能约束,设计坑道工程动荷段毁伤仿真实验训练样本约简分析机制,利用支持向量机结构风险最小化原则和非线性映射特性以及粒子群的快速全局优化特征,学习坑道工程动荷段毁伤仿真实验训练样本,快速建立训练样本的约简分类器,排除负类训练样本,为坑道动荷段毁伤评估提供优质的训练样本。算例表明,模型具有良好的收敛和精度。In damage simulation experiment on the active load section of protective engineering, training sets usually contain a large number of similar samples and redundant information to the classifiter model construction. A training sample reduction method was proposed for damage simulation experiment on the active load section of protective engineering based on improved support vector machine（SVM）. Following the reduction task and function constraints, training sample reduction analysis mechanism was designed for damage simulation experiment on the active load section of protective engineering. Based on the empirical or structural risk minimization principle and non-line relationship characteristics of SVM, and such quick global optimizing ability of particle swarm optimization（PSO）, a certain reduction classifiter was erected, which excludes negative training sample for damage simulation experiment on the active load section of protective engineering, providing excellent training sample for damage assessment operations. Case shows that the model has favorable efficiency and precision.

关 键 词：动荷段 毁伤 支持向量机 粒子群优化 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术] TP315[自动化与计算机技术—计算机科学与技术]