摘要
针对高炉炼铁过程的复杂性和时变性,以及观测值中包含异常值的问题,提出一种加权图的高炉过程故障检测方法以降低异常值的影响。采用基于图的突变点检测方法,考虑其具有无监督和非参数的优势。首先,根据加权的欧式距离利用最小生成树的方法得到连接图。然后,通过计算连接来自突变点前后两部分观测值的边数目作为统计量,实现故障检测。最后,利用数值仿真验证了算法的有效性,并在实际高炉过程中实施。结果表明,基于加权图方法,能够降低高炉过程中采集到数据矩阵中异常值的影响,提高故障检测的效果。
Since industrial processes are in general complex,proposing robust fault detection and identification is an important task to ensure process safety.In this paper a weight graph based fault detection method is proposed to reduce the influence of the outliers in blast furnace process.The introduced fault detection method has the advantage of being unsupervised and non-parametric.In this method,first the minimum spanning tree of observations is constructed.The weights are calculated by Euclidean distances,and a parameter is introduced to remove the outliers.Next the number of edges,which connect the two observations derived from two group,are counted to detect the fault.The power of proposed method was illustrated through numerical simulation of a blast furnace process.The results show that the weighted graph method can reduce the influence of outliers collected in the data matrix during the blast furnace process and improve the effect of fault detection.
作者
安汝峤
杨春节
潘怡君
AN Ru-qiao;YANG Chun-jie;PAN Yi-jun(Department of Control Science and Engineering,Zhejiang University,Hangzhou 310027,China;Shenyang Institute of Automation Chinese Academy of Sciences,Shenyang 110016,China)
出处
《高校化学工程学报》
EI
CAS
CSCD
北大核心
2020年第2期495-502,共8页
Journal of Chemical Engineering of Chinese Universities
基金
国家自然科学基金(61290321,61333007)。
关键词
过程控制
高炉
加权图
故障检测
process control
blast furnace
weight graph
fault detection
作者简介
安汝峤(1990-),男,黑龙江哈尔滨人,浙江大学博士生;通讯联系人:杨春节,E-mail:cjyang@iipc.zju.edu.cn。
