[1]吕照民,周革,苗晨.基于自适应主成分分析的化工过程在线监测[J].郑州大学学报(工学版),2020,41(01):44-48.[doi:10.13705/j.issn.1671-6833.2019.04.006]
 Lu Zhaomin,Zhou leather,Miao Chen.Online Monitoring of Chemical Process Based on Adaptive Principal Component Analysis[J].Journal of Zhengzhou University (Engineering Science),2020,41(01):44-48.[doi:10.13705/j.issn.1671-6833.2019.04.006]
点击复制

基于自适应主成分分析的化工过程在线监测()
分享到:

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
41
期数:
2020年01期
页码:
44-48
栏目:
出版日期:
2020-03-10

文章信息/Info

Title:
Online Monitoring of Chemical Process Based on Adaptive Principal Component Analysis
作者:
吕照民1周革2苗晨3
1. 上海工程技术大学城市轨道交通学院;2. 上海机电工程研究所;3. 上海工程技术大学电子电气工程学院
Author(s):
Lu Zhaomin 1Zhou leather 2Miao Chen 3
1. School of Urban Rail Transit, Shanghai University of Engineering Science; 2. Shanghai Institute of Mechanical and Electrical Engineering; 3. School of Electrical and Electronic Engineering, Shanghai University of Engineering Science
关键词:
过程监测主成分分析子空间自适应
Keywords:
process monitoringprincipal component analysissubspaceadaptive
DOI:
10.13705/j.issn.1671-6833.2019.04.006
文献标志码:
A
摘要:
当主成分分析(Principal Component Analysis, PCA)应用于过程监测时,不适当的成分选择方法会导致变异特征被分散或被淹没从而影响监测性能。针对这个问题本文提出了成分的自适应选择方法并用于过程监测,叫做自适应主成分分析(Adaptive Principal Component Analysis, APCA).自适应主成分应用于过程监测时主要包括三个步骤:1)首先,在离线建模时基于载荷矩阵通过欧氏距离计算各个成分的相似性,并基于每个成分选出与其相似性较高的成分构成多个成分子空间;2)其次,在线监测时基于在线样本的各成分通过核密度估计计算各个成分的变异概率,选择出变异概率最高的成分作为特征成分(Characteristic Component, CC);3)最后挑选出与CC对应的成分子空间,并构造统计量.通过数值仿真案例和田纳西伊斯曼(Tennessee Eastman, TE)过程证明了提出方法APCA的有效性。
Abstract:
When Principal Component Analysis (PCA) is applied to process monitoring, improper component selection method will cause variation characteristics to be dispersed or submerged, thus affecting monitoring performance. In order to solve this problem, this paper proposes an adaptive selection method of components and applies it to process monitoring, called Adaptive Principal Component Analysis (APCA). The application of adaptive principal components to process monitoring mainly includes three steps: 1 ) firstly, calculating the similarity of each component based on the load matrix through Euclidean distance during offline modeling, and selecting components with high similarity to each component to form multiple molecular spaces 2 ) Secondly, during on-line monitoring, the variation probability of each component is calculated by kernel density estimation based on each component of the on-line sample, and the component with the highest variation probability is selected as the characteristic component (CC) 3 ) Finally, the molecular space corresponding to CC is selected and statistics are constructed. Numerical simulation and Tennessee Eastman (TE) process prove the effectiveness of the proposed APCA

相似文献/References:

[1]董建华,王国胤,雍熙,等.基于Spark的标准化PCA算法[J].郑州大学学报(工学版),2017,38(05):7.[doi:10.13705/j.issn.1671-6833.2017.05.001]
 Dong Chee-hwa,Wang Guoyin,Yongxi,et al.Normalized PCA Algorithm Based on Spark[J].Journal of Zhengzhou University (Engineering Science),2017,38(01):7.[doi:10.13705/j.issn.1671-6833.2017.05.001]

更新日期/Last Update: 2020-02-22