一种基于模糊神经网络的故障诊断方法的研究

针对故障与征兆之间的复杂非线性映射关系，提出了一种新的诊断系统结构模型。用模糊神经网络从以往诊断实例的统计中得到诊断矩阵，用模糊变换法进行推理，通过动态权值综合法综合各规则的推理结论得出最后诊断结果，从而解决了规则不正交的问题，使诊断结果具有统计意义，且解决了常用方法中规则获取的困难和在使用中不易修改的问题。经对A320飞机故障实例的诊断仿真，得出了比较满意的结果。

On a New Method of Intelligent Fault Diagnosis

Taking into consideration the complex non-linear relationship between faults and symptoms, we propose a new structure model for diagnosis. We use fuzzy-neural network to learn the matrixes for diagnosis from the statistics of previous diagnostic instances, use fuzzy transform method to ratiocinate and use dynamic weight synthesis to integrate all ratiocinative results to get the diagnostic result. One of the important problems we discuss in detail in this paper is that one fault has many symptoms and one symptom may correspond to different faults, thus making the formulation of rules of diagnosis through learning difficult. Our detailed discussion and explanation reduce greatly the difficulty of such formulation of diagnostic rules. Thus, the acquisition and modification of rule will be much easier. The diagnosis process for an actual example of Airbus-320 is simulated. The simulation results show that the method is feasible and effective.