基于小样本RBF神经网络的导引头测高性能评估方法研究  

Research on high performance evaluation method of seeker based on small sample RBF neural network

作　　者：赵芃沛 赵甫 刘洋[3] 沈洋 方宗军 ZHAO Pengpei;ZHAO Fu;LIU Yang;SHEN Yang;FANG Zongjun(Shaanxi Aerospace Power Equipment Technology Co.,Ltd.,Xi’an 710025,China;Shaanxi Industrial Vocational and Technical College,Xianyang 712000,China;College of Aerospace,Northwestern Polytechnic University,Xi’an 710065,China)

机构地区：[1]陕西空天动力装备科技有限公司,西安710025 [2]陕西工业职业技术学院,陕西咸阳712000 [3]西北工业大学航天学院,西安710065

出　　处：《兵器装备工程学报》2023年第9期163-171,共9页Journal of Ordnance Equipment Engineering

摘　　要：针对导弹武器装备的试验成本高且试验数据少的问题,提出了一种小样本RBF神经网络模型。利用TOPSIS(逼近理想解排序法)法处理原始数据,充分挖掘数据深层信息,再利用Bootstrap法(自助法)对处理完的数据进行扩充,然后通过RBF神经网络(radial basis function)建立评估模型。最后,将小样本RBF神经网络模型应用于导引头测高性能评估以验证算法的有效性。仿真实验结果表明,小样本RBF神经网络模型的决定系数和误差系数相对于其他模型得到明显改善。该模型不仅规避了专家法、层次分析法、模糊综合评价法等常用效能评估方法主观性强的特点,同时,针对小样本问题提出了一种有效的解决方案。Aiming at the problem of lacking test data due to high test cost of missile weapon equipment,a small sample RBF neural network model is proposed.The original data is processed by TOPSIS method,the deep information of data is fully mined,the processed data is expanded by Bootstrap method,the evaluation model is established by radial basis function(RBF)neural network,and the small sample RBF neural network model is applied to the high-performance evaluation of seeker.The simulation results show that:the decision coefficient and error coefficient of small sample RBF neural network model are significantly improved compared with other models.The model not only avoids the subjectivity of expert method,analytic hierarchy process,and fuzzy comprehensive evaluation method,but also solves the problem of small sample size.

关 键 词：小样本 RBF神经网络 TOPSIS法 BOOTSTRAP法 性能评估 

分 类 号：TN06[电子电信—物理电子学]