基于改进证据理论和神经网络的故障诊断模型

Fault diagnosis model based on modified evidence theory and neural network

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。

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.The model could reduce the uncertainty of decision and greatly increase the precision of diagnosis.At last,the engine fault diagnosis example shows that the validity of the decision-level fusion fault diagnosis model.