基于集成Bootstrap方法的导弹试验鉴定精度综合评定  

Comprehensive Assessment of Accuracy of Missile in Test and Evaluation Based on Ensemble Bootstrap Method

作　　者：顾炎极 刘大鹏 任光 李成娟 李宗华 GU Yan-ji;LIU Da-peng;REN Guang;LI Cheng-juan;LI Zong-hua(China Aerospace Academy of Systems Science and Engineering,Information Control Institute,Beijing 100000,China)

机构地区：[1]中国航天系统科学与工程研究院信息控制研究所,北京100000

出　　处：《火力与指挥控制》2020年第7期177-181,共5页Fire Control & Command Control

摘　　要：针对导弹在试验鉴定过程中,试验样本极少,无法直接从极少的现场试验样本中得到可靠的精度评定结果的问题,使用仿真信息作为验前信息的基于集成Bootstrap方法对导弹精度进行综合评定,可以用于试验鉴定中精度综合评定。针对不平衡数据导致误差引入集成学习,即多组抽取样本和现场数据融合,通过多组弱置信度再得到强置信度的结果;分析了集成学习的误差必然会收敛于0;再由重要度抽样思想,采用了基于相容性水平的权重分配方案;最后采用了改进的总和规则作为集成规则,分析了参数误差。仿真结果表明基于集成Bootstrap方法能有效在小子样条件下进行导弹试验鉴定精度评定。Aiming at the problem of few missile test samples,it is impossible to obtain reliable accuracy evaluation results directly from few field test samples in the process of missile test and evaluation,the simulation information is used as the pre-test information based on the integrated Bootstrap method.The integrated learning is introduced to solve the error by unbalanced data,that is,multiple sampling groups and field data fusion,can obtain the results of strong confidence through multiple sets of weak confidence.The error of integrated learning will inevitably converge to zero.The importance sampling idea adopts the weight allocation scheme based on the compatibility level.Finally,the improved sum rule is adopted as the integrated rule,and the parameter error is analyzed.The simulation results show that the integrated Bootstrap method can effectively evaluate the accuracy under small sample condition.

关 键 词：导弹精度评定 BOOTSTRAP 方法 集成学习 验前信息 

分 类 号：TJ765.4[兵器科学与技术—武器系统与运用工程] TP391.9[自动化与计算机技术—计算机应用技术]