实现故障监控的智能预警
2020-08-13李虹韩永佳
李虹 韩永佳

摘 要: 本文提出了一种基于K-平均算法的智能监控预警算法。对于被监控的系统,该算法首先对所发生的故障进行分类,然后对故障类两两进行分析,建立故障实例之间的关联关系。在此基础上,该算法进一步搜索关联故障实例在故障类中的最佳分布并计算故障类之间的绝对和相对关联度。当关联度达到设定的阈值,我们则认为分组故障类存在关联关系,彼此之间存在相互触发动因。本文正是通过寻找这种关联关系来帮助实现故障监控的连带预警功能,实现潜在故障规避。为了对所提出的算法进行客观的评估,我们以企业内部的关键业务系统作为样本数据进行测试。测试结果表明,本算法能够高效并较准确地挖掘故障之间的关联性,对故障的智能监控预警具有实际意义。
关键词: K-平均算法;分类;故障;关联度;监控;预警
中图分类号: TP277 文献标识码: A DOI:10.3969/j.issn.1003-6970.2020.06.049
本文著录格式:李虹,韩永佳. 实现故障监控的智能预警[J]. 软件,2020,41(06):237241
【Abstract】: This paper presents an algorithm of intelligent pre-warning of exceptions based on K-means algorithm. For the system monitored, the algorithm firstly classifies the exceptions already occurred. Then, it works out the associate relationship of the exception instances for every two exception classifications and finds out the best distribution of the associated exception instances in the exception classifications. Thus, the algorithm calculates the absolute and relative association degree of every two exception classifications and chooses those exception classifications whose association degree meets the threshod request as the associated ones. Thus, the algorithm is able to pre-warn the exception of one classification possible to happen when it detects some exceptions of the associated classification already happened, and help to tell us to do something to avoid it. To evaluate the algorithm justly, we test it using some of our key business systems as data samples. The result shows that our algorithm can discover the relationship of exceptions efficiently and fairly exactly. It is quite valuable in the intelligent pre-warning of exceptions.
【Key words】: Data-mining; Classification; Exception; Association degree; Monitor; Pre-warning
0 引言
就如何提升客戶满意度,近年来业界展开了大量的探讨。需求满足、支撑到位是客户的基本诉求。而这全依赖于企业信息化水平的不断发展。业务系统的合理建设为各种业务的创新和规模发展提供了良好的基础。而后期系统的高效运维则是业务得以长久持续发展的保障。因此,业务系统支撑效率对客户的满意度起着至关重要的作用。因此,就如何提高业务系统健壮性(参见[1-6]),确保系统稳定可靠;加强系统监控(参见[7-10]),及时消除潜在故障,降低故障发生率;提升抢修技能、强化抢修工具配置,进而提高故障响应速度,降低故障处理时长均成为我们重点探讨的方向。……
