基于变量概率信息的因子分析监控方法

发布时间：2018-03-07 03:30

本文选题：因子分析　切入点：参数估计　出处：《化工学报》2017年07期 　论文类型：期刊论文

【摘要】：因子分析(factor analysis,FA)将噪声因素加入到建模过程中,可通过最大期望(expectation maximum,EM)算法建立模型。传统的FA(ST)指标仅利用了变量的期望信息而忽略了更能代表不确定性的方差信息,这可能会导致故障的漏报。通过对过程变量的概率分析,从本质上揭示了FA(ST)的这一缺陷。建模过程中的另一个重要因素是确定因子个数,使得在降维的同时能最大程度地保留对过程有用的信息。针对传统监控指标信息不足的问题,提出的负对数似然概率(negative log likelihood probability,NLLP)指标整合了更全面的概率信息;针对因子个数给定的问题,提出了一种整体-局部因子数确定法,使得因子和变量对于过程的信息解释率都达到收敛。最后通过数值例子和Tennessee Eastman(TE)过程验证了所提方法的有效性和优越性。
[Abstract]:Factor analysis factor analysis (FAA) adds the noise factor into the modeling process, and can be modeled by the maximum expectation maximization (EMN) algorithm. The traditional FAT index only uses the expected information of variables and neglects the variance information which can represent uncertainty. By analyzing the probability of process variables, the defect of FAT is revealed in essence. Another important factor in modeling is the number of factors. In order to reduce the dimension and keep the useful information to the maximum extent, the negative logarithmic likelihood probability negative log likelihood probability (NLLP) index integrates more comprehensive probability information in order to solve the problem of insufficient information of traditional monitoring index. For the problem of determining the number of factors, a global local factor number determination method is proposed. Both factors and variables converge to the information interpretation rate of the process. Finally, the effectiveness and superiority of the proposed method are verified by numerical examples and Tennessee Eastman (TET) process.
【作者单位】：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;
【基金】：国家自然科学基金项目(61374140,61673173) 中央高校基本科研业务费专项资金(222201714031) 中央高校基本科研业务费重点科研基地创新基金项目(222201717006)~~
【分类号】：O213

【相似文献】