论文部分内容阅读
对于复杂的工业过程,采集到的过程数据能反映出生产过程的内在变化和运行状况。本文提出一种新的多变量统计过程监测策略,数据建模过程包含主元分析(Principal Component Aanlysis,PCA)与正交局部保持投影(Orthogonal Locality PreservingProjection,OLPP)两步。首先利用PCA在不丢失任何信息的前提下将原始数据旋转成不相关的潜变量,然后再作OLPP以提取能表征过程正常数据内在局部近邻结构的特征用于故障检测。利用T~2和SPE(或Q)统计量以及核密度估计方法确定的控制限进行化工过程的在线监测,TE过程仿真实验验证了该混合方法的有效性和优越性。
For complex industrial processes, the collected process data can reflect the inherent changes in the production process and operating conditions. In this paper, a new multivariate statistical process monitoring strategy is proposed. The data modeling process includes two steps: Principal Component Analysis (PCA) and Orthogonal Locality Preserving Projection (OLPP). First, the PCA is used to rotate the original data into irrelevant latent variables without losing any information, and then OLPP is used to extract the features that can characterize the inherent local neighbors of the normal data of the process for fault detection. The online monitoring of chemical process was carried out by using the control limits determined by T ~ 2 and SPE (or Q) statistic and nuclear density estimation method. The TE process simulation experiment verified the effectiveness and superiority of the hybrid method.