Monitoring a paperboard machine using multivariate statistical process control

Abstract A principal component analysis (PCA) model was developed and used on-line to monitor paperboard manufacturing in a mill. Such monitoring is called multivariate statistical process control (MSPC). The mill information system tracks over 800 process variables. In consultation with operators, 177 variables were selected as relevant in monitoring a paperboard-manufacturing machine. Data for the model were selected according to a criteria function defined as the ideal process condition. The function states that production should be above a certain level and that all paperboard properties should remain within their specification limits for at least 7 h. The MSPC application is intended to monitor paperboard machine behaviour but not classify paperboard, so the model includes no variables pertaining to paperboard properties. Every minute, process variables are read and a prediction is made using the model. The result is plotted in a score plot, and a bar graph shows how each variable deviates from the model. The variables in the bar graph are sorted according to magnitude of deviation. When the model was introduced to operators, it was described as primarily being an intelligent system for sorting trend variables rather than as a multivariate application. After 6 months of operation, the PCA model came to be valued by operators on all six shift teams, as it facilitated detection of deviations and malfunctions in process equipment. The model can also be applied to the production of other grades of paperboard than the one for which it was designed.

[1]  Emmanuel G. Koukios,et al.  Fluorescence analysis of paper pulps , 1999 .

[2]  Age K. Smilde,et al.  Improved monitoring of batch processes by incorporating external information , 2002 .

[3]  A. J. Morris,et al.  An overview of multivariate statistical process control in continuous and batch process performance monitoring , 1996 .

[4]  A. J. Morris,et al.  Performance monitoring of a multi-product semi-batch process , 2001 .

[5]  O. Morgenthaler,et al.  Proceedings of the Conference , 1930 .

[6]  A. J. Morris,et al.  Monitoring the performance of the paper making process , 1999 .

[7]  Graham C. Goodwin,et al.  Predicting the performance of soft sensors as a route to low cost automation , 2000 .

[8]  H. Budman,et al.  BOD5 estimation for pulp and paper mill effluent using UV absorbance. , 2001, Water research.

[9]  L. Danielsson,et al.  Quantitative in-line monitoring of powder blending by near infrared reflection spectroscopy , 2002 .

[10]  Oscar A.Z. Sotomayor,et al.  Analysis of activated sludge process using multivariate statistical tools—a PCA approach , 2002 .

[11]  Kim H. Esbensen,et al.  Non-invasive monitoring of powder breakage during pneumatic transportation using acoustic chemometrics , 2003 .

[12]  John F. MacGregor,et al.  Multivariate SPC charts for monitoring batch processes , 1995 .

[13]  Leis J.M,et al.  Proceedings of the Symposium , 1997 .

[14]  Carl-Fredrik Mandenius,et al.  Comparison Between Linear and Nonlinear Prediction Models for Monitoring of a Paperboard Machine , 2002 .