基于时间序列的分布式网络节点故障检测系统设计

Design of distributed network node fault detection system based on time series

目前的分布式网络节点故障检测系统对节点的异常度查询不足,容易产生误判,影响故障检测系统的准确度。因此,文中提出基于时间序列的分布式网络节点故障检测系统设计。首先,对系统连接的电源电路进行设计,在系统与网络连接中,选择TL718协议转换芯片,并设计网络脉冲测试器。在系统的软件设计上,运用小波分解,将脉冲信号的空间和频率进行变化,提取出信号异常信息,运用时间序列法对异常信号进行异常度查询,最后对异常程度较高的节点计算异常密度函数,判断是否存在故障。为了验证设计系统的可行性,设计相关实验,并使用传统系统与所设计系统对某公司的分布式网络中的节点进行故障检测,实验结果证明,设计的分布式网络节点故障检测系统具有较高的准确度,设计系统具有一定可行性。

The current distributed network node fault detection system has no sufficient query against the network node anomaly degree,so it is possible to produce misjudgement and affects the accuracy of fault detection.Therefore,a distributed network node fault detection system based on time series is proposed.The power supply circuit of the system is designed.In the connection between the system and the network,TL718 protocol conversion chip is selected as the principal part,and a network pulse tester is designed.In the software design of the system,the wavelet decomposition is used to change the space and frequency of the pulse signal,extract the abnormal information of the signal,use the time series method to query the abnormal degree of the abnormal signal,and calculate the abnormal density function of the node with higher abnormal degree to judge whether there is a fault.In order to verify the feasibility of the designed system,a relevant experiment was designed,in which the traditional system and the designed system were adopted to detect the fault of nodes in a company's distributed network.The experimental results show that the distributed network node fault detection system designed in this paper has high accuracy and feasibility.