基于混合分块DMICA-PCA的全流程过程监控方法

发布时间：2018-04-18 16:47

本文选题：主元分析 + 过程控制　；参考：《化工学报》2017年02期

【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。
[Abstract]:Block strategy is widely used in the field of whole process process monitoring to solve the problem of high complexity of variable relationship in the whole process process, but the traditional block strategy and subblock modeling method do not consider the dynamic problem of the process.The traditional partitioning strategy relies on process knowledge or process data information, which affects the effect of process monitoring. Therefore, a process monitoring method based on mixed block DMICA-PCA is proposed.After analyzing the dynamic nature of the process, the known partial process knowledge is used to divide the variables into blocks, and then the improved generalized Dice's coefficients between the blocks are used to further divide the variables.Then the DMICA-PCA method is used to model each sub-block to obtain the statistics of the sub-block, and the total joint index is obtained by weighted method for fault detection.At the same time, an improved fault diagnosis method is adopted for each sub-block, which improves the diagnosis effect.Finally, the method is applied to the process monitoring of te process, and the effectiveness of the method is proved.
【作者单位】：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;上海交通大学电工与电子技术中心;
【基金】：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)~~
【分类号】：TP277

【相似文献】