CAJ | 학술논문

传统基于主元分析的故障检测方法大多假设工业过程只运行在1个稳定工况,数据服从单一的高斯分布.若这些方法直接用于多工况过程则将会产生大量的误检.为此,本文提出了1种基于高斯混合模型的多工况过程监测方法.首先利用PCA变换对过程数据集进行降维,在主元空间建立高斯混合模型对过程数据进行聚类,自动获取工况数和相关分布特性.然后对每个工况建立主元分析(principal component analysis,PCA)模型来描述整个运行过程数据分布的统计特性.最后在过程监测中,根据监测样本属于各个工况的概率构造综合统计量,实现对多工况过程的故障检测.TE过程的仿真结果表明,本文提出的方法与传统的PCA方法相比,能自动获取工况和精确估计各个工况的统计特性,从而能更准确及时地检测出多工况过程的各种故障.
전통기우주원분석적고장검측방법대다가설공업과정지운행재1개은정공황,수거복종단일적고사분포.약저사방법직접용우다공황과정칙장회산생대량적오검.위차,본문제출료1충기우고사혼합모형적다공황과정감측방법.수선이용PCA변환대과정수거집진행강유,재주원공간건립고사혼합모형대과정수거진행취류,자동획취공황수화상관분포특성.연후대매개공황건립주원분석(principal component analysis,PCA)모형래묘술정개운행과정수거분포적통계특성.최후재과정감측중,근거감측양본속우각개공황적개솔구조종합통계량,실현대다공황과정적고장검측.TE과정적방진결과표명,본문제출적방법여전통적PCA방법상비,능자동획취공황화정학고계각개공황적통계특성,종이능경준학급시지검측출다공황과정적각충고장.
Traditional fault detection methods based on pinciple component analysis(PCA) rely on the assumption that the process has one nominal operating region and the operating data follow a unimodal Gaussian distribution.The application of these approaches to an industrial process with multiple operating modes would always trigger false alarms.Thus,a new multimode process monitoring approach based on Gaussian mixture models(GMM) is proposed in this paper.First a GMM is constructed in the model subspace obtained by PCA transformation to characterize the multiple operating regions,each of which corresponds to a Gaussian component.Then a principal component model is established for individual operating mode to describle the statistical features of whole operating process.Finally,an overall statistics chart,according to the posterior probability of a monitored sample belong to each Gaussian component,is defined to monitoring the multimode process.The validity and effectiveness of the proposed monitoring approach is illustrated by the Tennessee-Eastman challenge process.The comparison of monitoring results demonstrated that the proposed approach can achive a good parameters estimation of the multiple operating regions with automatically select the number of modes.Therefore,it is superior to the conventional PCA method and can achieve accurate and early detection of various types of faults in mulitimode processes.