CIESC Journal ›› 2017, Vol. 68 ›› Issue (2): 759-766.DOI: 10.11949/j.issn.0438-1157.20161309

Previous Articles     Next Articles

Plant-wide process monitoring based on mixed multiblock DMICA-PCA

JIANG Wei1, WANG Zhenlei1, WANG Xin2   

  1. 1. State Key Laboratory of Chemical Engineering, Key Laboratory of Advanced Control and Optimization for Chemical Processes, East China University of Science and Technology, Shanghai 200237, China;
    2. Center of Electrical & Electronic Technology, Shanghai Jiao Tong University, Shanghai 200240, China
  • Received:2016-09-19 Revised:2016-12-05 Online:2017-02-05 Published:2017-02-05
  • Supported by:

    supported by the Key Program of National Natural Science Foundation of China (61134007), the Youth Program of National Natural Science Foundation of China (61403141), the Shanghai "Technology Innovation Action Plan" Development Platform for Building Projects (13DZ2295300), the Natural Science Foundation of Shanghai (14ZR1421800) and the State Key Laboratory of Synthetical Automation for Process Industries (PAL-N201404).

基于混合分块DMICA-PCA的全流程过程监控方法

江伟1, 王振雷1, 王昕2   

  1. 1. 化学工程联合国家重点实验室, 华东理工大学化工过程先进控制和优化技术教育部重点实验室, 上海 200237;
    2. 上海交通大学电工与电子技术中心, 上海 200240
  • 通讯作者: 王振雷
  • 基金资助:

    国家自然科学基金重点项目(61134007);国家自然科学基金青年项目(61403141);上海市“科技创新行动计划”研发平台建设项目(13DZ2295300);上海市自然科学基金项目(14ZR1421800);流程工业综合自动化国家重点实验室开放课题基金资助项目(PAL-N201404)。

Abstract:

Multiblock strategy is widely used in plant-wide process monitoring to solve problems with complicated relationships between process variables. Traditional multiblock strategies and sub-block modeling methods are not effective in plant-wide process monitoring, because dynamic characteristics of the process have not been considered and knowledge or data information of the process is exclusively exploited. A mixed multiblock DMICA-PCA method was proposed to improve process monitoring performance. First, variables were sliced into initial sub-blocks by obtained process knowledge after analysis of process dynamics and further sliced into final sub-blocks by modified general Dice's coefficient (MGDC) between variables of initial sub-blocks. Then, the DMICA-PCA method was used to establish model and acquire statistical values of variables in final sub-blocks and a combined overall index from weighted sum was developed for fault detection, which improved performances by simultaneous diagnosis on each sub-block. Effectiveness of the proposed method was validated on monitoring the Tennessee-Eastman (TE) process.

Key words: principal component analysis, process control, process systems, mixed multiblock, plant-wide process, modified general Dice's coefficient

摘要:

分块策略被广泛运用于全流程过程监控领域,以解决全流程过程变量关系复杂性较高的问题,但传统的分块策略与子块建模方法都未考虑过程的动态性问题,并且传统的分块策略都片面依赖于过程知识或过程数据信息,影响了过程监控的效果,为此提出了一种基于混合分块DMICA-PCA的过程监控方法。在分析过程的动态性后,先利用已知的部分过程知识进行变量的初步分块,接着利用各分块变量之间改进的广义Dice's系数(MGDC)进行进一步的分块。然后采用DMICA-PCA方法对每个子块进行建模得到子块的统计量,并通过加权方法得到总的联合指标进行故障检测。同时对每个子块采用改进的故障诊断方法,提高了诊断效果。最后将该方法应用在TE过程的过程监控中,证明了该方法的有效性。

关键词: 主元分析, 过程控制, 过程系统, 混合分块, 全流程, 改进的广义Dice's系数

CLC Number: