基于多状态贝叶斯网络的导弹质量状态评估

针对“是非制”导弹质量状态分类方法过于粗放、信息资源利用不充分的问题，提出了基于DS证据理论-层次分析法的条件概率赋值方法，并构建了贝叶斯网络的质量状态评估模型。在分析性能特征参数的基础上，将导弹状态划分为良好、较好、堪用、拟故障和故障5个等级。对单位不一、阈值各异的测试参数进行归一化处理，以改进的岭形分布函数确定各质量状态等级的隶属度，构建基于贝叶斯网络的导弹质量状态评估模型。鉴于各节点间为非确定性逻辑关系、数据信息相对缺乏，提出基于DS证据理论-层次分析法的多状态贝叶斯网络条件概率值确定方法。以某型导弹为例，经过验证，DS证据理论-层次分析法确定条件概率值的方法将不确定度降低到了5%左右，并通过横向对比与纵向对比确定了该模型的合理性与可信性。

Missile Condition Assessment Based on Multi-state Bayesian Network

A condition assessment model of missile based on Bayesian network (BN) with conditional probability table (CPT) which is determined by using DS evidence theory and analytic hierarchy process (AHP) is established for the extensive way and inadequate information resources ultilization of traditional state classification method.Missile state is divided into five grades,i.e.,perfect,better,usable,pseudo-fault and fault,based on the analysis of characteristic parameters.Normalization processing is carried out on the test parameters of different units and threshold values,and the modified ridge shape functions are used to determine the degree of membership of all the states.A BN-based condition assessment model of a missile is established.In the view of uncertain logical relationship between nodes and lacking of data,a method is proposed to determine CPT of multi-state Bayesian network on the basis of DS evidence theory and AHP.The result shows that uncertainty is decreased to about five percent by utilizing DS/AHP method to determine CPT.The reasonability and credibility of BN model are verified through horizontal and vertical comparison.