变精度粗糙集与神经网络在故障诊断中的应用

针对标准的粗糙集理论不能很好地处理带有噪声的数据,而故障诊断信息中难以避免地存在噪声数据,对此,提出了SOM网络-变精度粗糙集-RBF神经网络的故障诊断方法：首先应用SOM网络对故障诊断数据中的连续属性值进行离散化,然后利用变精度粗糙集理论的属性依赖度进行启发式约简,据此得到最优决策系统,最后在最优决策系统的基础上设计RBF神经网络进行故障诊断。实例验证了该方法的可行性,且故障诊断正确率高。

Application of Variable Precision Rough Set and Neural Network to Fault Diagnosis

Considering that the standard rough set theory cannot effectively process the noise data,in addition,there were always noise data in fault diagnosis data,a new method of SOM network-variable precision rough set-RBF neural network for fault diagnosis was proposed.Firstly,the continuous attributes in diagnostic decision system were discretized with SOM network.Then,reducts were found based on attribute dependence of variable precision rough set theory,and the optimal diagnostic decision was determined.Finally,according to the optimal decision system,RBF neural network was designed for fault diagnosis.A practical example was given to show the method is feasible and available,with high rate of accurate fault diagnosis obtained.