面向互联网的局域底层软件运行故障检测仿真

Internet-Oriented Local Area Underlying Software Operation Fault Detection Simulation

针对基于主成分分析的故障检测方法只考虑软件运行数据集非高斯信息的提取而没有考虑高阶特征,导致故障检测延迟时间较长,检测率低的问题,提出一种面向互联网的局域底层软件运行故障检测方法。采用贝叶斯推断方法将局域底层软件运行过程变量和质量变量间的关联关系的互信息融合,选择含有质量变量信息量最大的一组过程变量,利用主成分分析方法对这组过程变量统计建模,依据加权互信息选取故障低阶信息量最大的特征,构建故障低阶数据统计量,采用统计模量分析方法将故障低阶特征映射到高阶特征空间中,统计故障高阶特征数据,构成运行周期内的故障高阶数据统计量,将此周期内统计模量作为行向量建立训练集,应用核主元分析方法对训练集进行故障检测同时确定控制限,弥补了主成分分析方法无法提取故障高阶特征的弊端。实验结果表明,所提方法能够有效提升故障的检测时间,提高故障检测率。

This article proposed a method for detecting fault of local underlying software area on internet. This method used Bayesian inference method to fuse the mutual information of association relation between process varia- bles and quality variables in the operation of local underlying software, and then selected a set of process variables with the maximum amount of information of quality variables to carry out statistical modeling by using principal com- ponent analysis. Based on weighted mutual information, our research selected the feature with the maximum amount of low-order information to construct the low-order fault data statistic. Moreover, the research used statistical modu- lus analysis to map the low-order feature of fault into the high-order feature space. Meanwhile, the research counted the high-order feature data of fault to constitute high order data statistics of fault during operational cycle time. In ad- dition, statistical modulus in this period was regarded as the row vector to establish the training set. Finally, we used the kernel principal component analysis method to detect the fault in training set and determine the control limit at the same time. Simulation results show that the proposed method can effectively improve the fault detection time and the fault detection rate.