【摘要】 针对过程中难以检测到的微小、缓变故障的检测问题,以及过程中普遍存在的非高斯信息,提出一种新的多变量统计过程监测方法.把传统的单变量累计和控制图(CUSUM)扩展为多变量的形式,并与独立成分分析(ICA)和传统的主元分析(PCA)方法相结合,构成新的MCUSUM-ICA-PCA方法,采用ICA-PCA两步信息提取策略,完整地提取出过程的非高斯和高斯信息,重新构造统计量并建立其对应的统计限.通过对Tennessee Eastman(TE)过程的仿真研究,验证了该方法的可行性和有效性,改善了该过程微小故障的检测效果,从而更好地保证过程运行的安全、稳定性.更多还原

【Abstract】 In order to detect the process influenced by gradual small shifts and extract the non-Gaussian information,a new multivariate statistical process monitoring method was proposed,which extended the conventional cumulative sum(CUSUM) to MCUSUM and combined it with independent component analysis(ICA) and principal component analysis(PCA) to form MCUSUM-ICA-PCA.A two-step information extraction strategy was proposed to extract the full information(including non-Gaussian and Gaussian),then three statistics and their corresponding statistical limits were built.A case study of the Tennessee Eastman(TE) process shows that the proposed method is efficient,the process monitoring performance is evidently improved,and it also enhances the reliability and stability of the TE process.更多还原