摘要
根据操作流程,提出了构建复杂武器系统健康管理指标体系的两种方法:层析法和聚合法,并对其优缺点进行了分析;通过对BP网络分析,得出网络权值矩阵能够反映输入对输出的影响程度的结论,并基于此原理进行了健康管理指标筛选;利用BP网络良好的非线性映射能力,提出了一种构造隶属度矩阵的客观方法,并对指标体系进行了评价;经实际验证,样本训练均方误差为9.887×10-4,构造的隶属度矩阵能够较好反映各级别之间的隶属关系,构造的健康管理指标体系比较合理。
According to operation process, two constructing methods for health management index system of complex weapon system are brought forward: layer analyzing method and aggregation method, the advantage and the disadvantage of it are analyzed. Through analyzing BP network, the conclusion that weight matrix could reflect influence degree of input to the output is put forward, and based on this, the health management index is filtrated. An objective method that constructing membership grade matrix is put forward based on better nonlinear mapped ability of BP network, and the index system is evaluated. By actual validation, mean square error of the training sample is 9. 887× 10-4 , the membership grade matrix could reflect the relationship of each class preferably, and the health management index system is more reasonable.
出处
《计算机测量与控制》
北大核心
2014年第4期1169-1171,1192,共4页
Computer Measurement &Control
关键词
武器系统
健康管理
指标体系
BP网络
weapon system
health management, index system
BP network