基于改进证据理论和神经网络的故障诊断模型*

针对单一故障诊断方法精度低的问题，提出了一种基于D-S证据理论和神经网络相融合的决策层融合故障诊断模型。该方法利用证据理论来处理不精确的、模糊的信息，用神经网络来处理证据理论中的基本可信度分配问题。由于证据理论合成公式无法处理高冲突的证据，提出了一种改进的基于冲突焦元的证据合成规则。该模型在降低决策不确定性的同时大大提高了诊断的精度。最后通过发动机故障诊断实例验证了该模型的有效性。

Fault diagnosis model based on modified evidence theory and neural network

Directing to the low precision of single fault diagnosis systerm,this paper put forward the decision-level fusion fault diagnosis model which fusing neural network and D-S evidence.The method used D-S's evidence to deal with inaccuracy and fuzzy information,and evidence's basic belief assignment could be sloved by neural network.Proposed a new combination rule,which based on reallocation of the basic probability assigned to conflict focal elements.The method could solve the problem of conflicting evidences....