Abstract: This paper introduces rough sets theory.And rough sets theory is used to eliminate unnecessary attributes from the decision table.Then make improvement BP network as the back processor of rough set，and a fault diagnosis model based on rough set and neural network.The result of emluator indicats that this method can reduce the needed training samples and simply the neural network structure and shortened the training time of the network.It is estimated that the optimized strategy may be further applied in fault diagnoses.

Key words: fault diagnosis, neural network, rough set, attribute reduction

摘要： 通过引入粗糙集理论，利用可辨识矩阵约简算法对故障诊断决策表进行属性约简，剔除其中不必要的属性，然后构造改进的BP神经网络作为粗糙集的后端处理机，构造了基于粗糙集与神经网络的故障诊断模型。仿真结果表明，该方法可以有效地减少输入层个数，简化神经网络结构，减少网络的训练时间，在故障诊断中有良好的应用前景。

关键词: 故障诊断, 神经网络, 粗糙集, 属性约简