基于PCA权重的化工报警阈值优化
发布时间：2018-08-03 11:19
【摘要】：为避免海量报警影响正常操作,对报警阈值进行了优化.考虑了变量相关性,利用主元分析法(PCA)分析变量的重要程度,并求出变量权重;结合国际标准,根据变量权重分配变量报警数,优化报警阈值;在平行坐标图上将变量数据及报警阈值可视化,查看正常和异常工况区域,考察各变量之间的变化趋势,提取有效报警值;对某工业原油常减压操作数据进行验证,结果表明优化方法能有效减少报警数.
[Abstract]:In order to avoid massive alarm affecting normal operation, the alarm threshold is optimized. Considering the correlation of variables, the importance of variables is analyzed by principal component analysis method (PCA), and the weight of variables is calculated, and the alarm threshold is optimized according to the international standard. The variable data and alarm threshold are visualized on the parallel coordinate graph, the normal and abnormal working condition regions are inspected, the variation trend among the variables is investigated, the effective alarm value is extracted, and the operating data of a certain industrial crude oil is verified. The results show that the optimization method can effectively reduce the number of alarms.
【作者单位】： 青岛科技大学化工学院;
【基金】：国家自然科学基金资助项目(编号:21576143) 山东省自然科学基金资助项目(编号:ZR2013BL008)
【分类号】：TE687;TP277
，
本文编号：2161584
[Abstract]:In order to avoid massive alarm affecting normal operation, the alarm threshold is optimized. Considering the correlation of variables, the importance of variables is analyzed by principal component analysis method (PCA), and the weight of variables is calculated, and the alarm threshold is optimized according to the international standard. The variable data and alarm threshold are visualized on the parallel coordinate graph, the normal and abnormal working condition regions are inspected, the variation trend among the variables is investigated, the effective alarm value is extracted, and the operating data of a certain industrial crude oil is verified. The results show that the optimization method can effectively reduce the number of alarms.
【作者单位】： 青岛科技大学化工学院;
【基金】：国家自然科学基金资助项目(编号:21576143) 山东省自然科学基金资助项目(编号:ZR2013BL008)
【分类号】：TE687;TP277
，
本文编号：2161584
本文链接：https://www.wllwen.com/kejilunwen/anquangongcheng/2161584.html
