基于粗糙集-神经网络集成的故障诊断  

Fault Diagnosis Based on Rough Set and Neural Network Ensemble

作　　者：郭小荟[1] 马小平[1] 

机构地区：[1]中国矿业大学信息与电气工程学院

出　　处：《控制工程》2007年第1期53-56,共4页Control Engineering of China

基　　金：江苏省自然科学基金资助项目(BK2003026)

摘　　要：综合粗糙集和神经网络的优点,提出一种基于粗糙集-神经网络集成的智能故障诊断模型。在数据采集和预处理的基础上,利用粗糙集(RS)理论对原始故障诊断样本进行离散化处理,并根据条件属性(集)对决策属性的正域的大小来选择属性,提取出对诊断故障贡献最大的最小故障特征子集,从而确定神经网络的拓扑结构;通过网络训练建立故障特征与故障之间的映射关系,采用神经网络集成的方法实现故障的诊断。通过热电厂发电机组的故障诊断实例,表明了这种故障诊断方法的工程有效性。An intelligent fault diagnosis model using rough set and artificial neural network ensemble is developed, considering that neural networks are best for solving nonlinear problems while rough set is good for data reduction. Basis on data acquisition and pretreatment, the original fault diagnosis samples are discretized by using rough set theory. According to the decision attribute positive region of condition attribute （s）, the minimum fault feature subset is selected, and the neural network structure is detenninted. The networks are trained to reflect the mapping between inputs and outputs, and network ensemble is used to realize the fault diagnosis. An example shows the effectiveness of the method.

关 键 词：故障诊断 粗糙集 神经网络 神经网络集成 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]