Abstract:

Rough sets theory combined with Neural Network was applied to intelligent diagnosis system, based on the importance of attribute and the consistency of decision table, SOM(self-organizing map) neural network was employed to discretize continuous data. Then rough sets theory was applied to reduce attribute and extract the primary feature which will be the foundation of structuring the sub-network. The final conclusions are reached by combing the results of sub-networks based on D-S(dempster-shafer) evidence theory. The method was applied to diagnosis the aero engine wear fault.Example shows the validity of the method proposed.

Key words: wear fault, aero engine, rough set, integrated neural network

摘要：

将粗糙集理论和神经网络相结合并应用到航空发动机磨损故障诊断中，依据属性的重要性和决策表的相容性，用自组织神经网络完成连续数据离散处理这一关键环节，采用粗糙集理论对征兆信息进行属性约简，获取征兆的主要特征，为神经网络结构简化和子神经网络的构成等奠定了基础，通过基于D-S证据理论的方法得到最终的融合结果。将该方法用于某型航空发动机的磨损故障诊断专家系统中，实验证明了该方法的有效性。

关键词: 磨损故障, 航空发动机, 粗糙集, 集成神经网络