粗糙集CMAC神经网络故障诊断策略

Rough Set-Based CMAC Neural Network for Fault Diagrosis

提出一种基于粗糙集CMAC神经网络的智能互补融合的诊断策略 .该策略利用粗糙集理论对数据样本进行数据浓缩 ,提取初步的诊断规则 .对初步的诊断规则通过神经网络进行粗映射 ,利用神经网络的分类逼近能力 ,建立故障状态空间到诊断空间的精确映射 .大大提高了神经网络的收敛速度和逼近精度 .将该神经网络应用于的变压器故障诊断实例 ,结果表明 ,该神经网络具有分类逼近能力强 ,计算量小等优点 .

A rough set based CMAC neural network is put forward as intelligent complementary and blending tactics of diagnosis. This tactics carry out data compaction on data samples and extract initial diagnostic rule by using rough set theory. To carry out rough mapping on the initial diagnostic rule through neural network and to use the sort approximation ability of neural network, an exact mapping from space of fault state to space of diagnosis is established by which convergence rate and approximation accuracy are greatly improved. This neural network is applied to the example of fault diagnosis of transformer. The result shows that the neural network is strong in sort approximation ability and small in workload of computation and high in rate of correct diagnosis, as compared with that of conventional neural network.