论文信息 - Process monitoring based on canonical variate analysis

Process monitoring based on canonical variate analysis

A new process monitoring strategy is proposed based on canonical variate analysis (CVA). State-space models are identified from input/output data collected over different time periods, and then an optimal test of hypothesis concerning process change is computed by comparing the respective model fits, using the Akaike Information Criterion. The new process monitoring strategy is illustrated in a simulation example, a continuous tank reactor (CSTR) problem.

Dale E. Seborg | Yi Wang | Wallace E. Larimore

[1] W. Harmon Ray,et al. Chemometric methods for process monitoring and high‐performance controller design , 1992 .

[2] Ali Cinar,et al. Statistical process monitoring and disturbance diagnosis in multivariable continuous processes , 1996 .

[3] Dale E. Seborg,et al. Identification of chemical processes using canonical variate analysis , 1994 .

[4] T. McAvoy,et al. Nonlinear principal component analysis—Based on principal curves and neural networks , 1996 .

[5] Duncan A. Mellichamp,et al. Identification of chemical processes using canonical variate analysis , 1990, 29th IEEE Conference on Decision and Control.

[6] Dale E. Seborg,et al. Process Monitoring using Canonical Variate Analysis and Principal Component Analysis , 1997 .

[7] Thomas E. Marlin,et al. Multivariate statistical monitoring of process operating performance , 1991 .

[8] Dong Dong,et al. Nonlinear principal component analysis-based on principal curves and neural networks , 1994, Proceedings of 1994 American Control Conference - ACC '94.

[9] Douglas C. Montgomery,et al. Introduction to Statistical Quality Control , 1986 .

[10] Theodora Kourti,et al. Multivariate SPC Methods for Process and Product Monitoring , 1996 .