摘要
针对电动机状态监测中异常数据检测存在的准确度低和测量值不精确等问题,文章提出一种基于区间数的不确定数据流异常检测算法。在该算法中,首先引入区间数的方法描述电动机状态监测中测量信号的不确定性,再利用区间数的位置关系对当前窗口的不确定数据进行剪枝,去除大多数正常的数据,最后根据距离值重新排列当前窗口的数据,收缩数据点的K-最近邻对象的查询范围。实验结果表明,该方法具有较高的检测精度和较低的计算复杂度,并能很好的运用到电动机状态监测中的异常数据检测。
Aiming at the inaccuracy of outlier data and measured data in the control system of motor,an outlier detection algorithm based on interval data was proposed in uncertain data stream. In this algorithm,firstly,the interval data introduced was used to express the inaccuracy of measured signal; Then all the data in the sliding windows was pruned by the position between two interval data,to dislodge most of the normal data; Finally,the current windowdata based on their distance with the first data was reordered,to reduce query data in k-close distance. Experiment results showed that the algorithm not only possessed better clustering precision with lowcomputing complexity but also applied to outlier data detection in the motor monitoring system well.
出处
《组合机床与自动化加工技术》
北大核心
2016年第8期34-38,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
国家科技支撑计划(2012BAF12B14)
贵州省重大科技专项(黔科合重大专项字(2012)6018)
贵州省科学技术基金项目(黔科合J字[2011]2196号)
贵州省工业攻关项目(黔科合GY字(2013)3020)