多属性加权模糊贝叶斯的复杂网络故障自修复技术  

Complex network fault self-repair mechanism with multi-attribute weighted fuzzy Bayesian

作　　者：蒋勇[1,2] 赵作鹏[2] 

机构地区：[1]江苏联合职业技术学院信息技术系,江苏徐州221008 [2]中国矿业大学计算机科学与技术学院,江苏徐州221008

出　　处：《计算机应用研究》2015年第8期2378-2381,共4页Application Research of Computers

基　　金：江苏省自然科学基金资助项目(BK2012129)

摘　　要：为了提高对复杂网络进行故障诊断时的准确率,以及实现故障节点的有效自修复,提出一种多属性加权模糊贝叶斯的复杂网络故障自修复机制。建立贝叶斯网络结构模型,针对故障网络和故障节点进行条件概率估计,实现故障类别诊断。在该模型的基础上引入了多属性值和模糊集合理论进行扩展,提出了一种多属性加权模糊贝叶斯,提升模型对节点进行故障诊断时的灵敏度和准确度。对网络和节点进行故障诊断后,采用网络故障自修复机制,在查找出节点故障类型,采取有效的能量分配方法来修复节点。实验仿真及对比表明,该方法相比基于神经网络的故障诊断方法、基于半监督聚类的故障诊断方法以及基于隐马尔可夫模型的故障诊断方法具有更好的故障诊断和修复性能。In order to improve the accuracy of complex network fault diagnosis, and achieve effective self-repair the faited node, this paper proposed a complex network fault self-repair mechanism with multi-attribute weighted fuzzy Bayesian. It built Bayesian network structure model for the conditional probability estimated for faults and fault network nodes to achieve fault diagnosis category. On the basis of the model on the introduction of multi-attribute values and fuzzy set theory to expand, it proposed a multi-attribute weighted fuzzy Bayesian model to enhance the sensitivity and accuracy of the nodes for fauh diagno- sis. After the nodes of the network and troubleshoot network problems using self-repair mechanism, to find out the type of node failures and adopt effective energy allocation method to repair nodes. The simulation and comparison show that this method compared to the fault diagnosis method based on neural network fault diagnosis method, based on semi-supervised clustering and fault diagnosis method, and based on hidden Markov model has better performance troubleshooting and repair.

关 键 词：故障诊断 故障修复 贝叶斯 多属性加权模糊贝叶斯 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]