基于一种新型粗糙集神经网络的故障诊断*

分析了粗糙集与神经网络各自的优缺点，结合粗糙集与神经网络提出了一种基于SOFM网络的新型粗糙集神经网络，给出了该网络的流程图，描述了系统各组成部分的工作原理。应用一实例验证了该网络在故障诊断中的有效性。结果表明，新网络较好地解决了训练样本的大小、样本质量等对人工神经网络的精度以及泛化能力有直接影响的问题，简化了神经网络的结构，缩短了训练时间，并实现了结果可视化,最后得到了故障可视拓扑映射图。

Study of Fault Diagnosis Based on New Rough Set Neural Networks

A new rough set-neural network based on Self Organized Feature Map(SOFM) network is presented in this paper.Rough set theory has powerful capability for qualitative analysis,while SOFM network can approach most problems fast and visualize result easily.By combing those advantages of the two theories,a system flow chart is designed and the work principles of each part are described.The validity of these models is tested by practical examples.The method can decrease the computation time,and visualize the result.Finally,the visualized topology-ordering map is obtained.