一种基于PMC模型下的概率性矩阵诊断算法  

Probability matrix diagnosis algorithm based on PMC model

作　　者：冯海林[1] 雷花[1] 梁伦 

机构地区：[1]西安电子科技大学数学与统计学院,陕西西安710126

出　　处：《南京理工大学学报》2017年第4期479-485,共7页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(71271165)

摘　　要：系统级故障诊断是提高多处理器系统可靠性的必要手段。为了有效定位多处理系统中的故障单元,该文建立了一种基于PMC模型t可诊断条件下的概率性矩阵诊断算法。首先对一般概率性矩阵诊断算法进行仿真分析获悉其具有较高的误检率,在诊断过程中引进绝对故障基和节点集团思想,通过计算绝对故障基以寻找系统中的部分故障处理机,集团用于将不确定状态的节点单元分类以补充正常节点集合,改善了原诊断的限制条件。仿真实验验证:改进后的概率性矩阵诊断算法保持了很高的检测精度,并且随着节点数的增多极大地降低了误检率,提高了诊断效果,使得该算法具有广泛的适用性。The system-level fault diagnosis, an efficient method, is an essential subject for the expanding multiprocessor system. In order to maintain the proper functioning of the system via locating or evading the fault nodes,the probability matrix diagnosis algorithm is studied under the PMC model in t diagnosable system. Firstly, according to the analysis result of the simulation experiments on the general probablility matrix diagnostic, the higher fault alarm rate is presented. The absolute fault nodes aggregation based on the syndrome matrix is introduced to identify some fault nodes,the nodes grouping is used to replenish the non-fault sets, and the rigorous condition is impaired. Finally, the modifiedprobability matrix diagnosis algorithm is proposed to improve the diagnostic efficiency. Simulation experiments show that it keeps the superiority of high detection accuracy, reduces fault alarm rate with the nodes increasing,and confirms the impressive diagnostic efficiency and extensive application.

关 键 词：系统级故障诊断 概率性矩阵诊断算法 绝对故障基 节点集团 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]