UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Prediction and prevention of sheet break using partial least squares and an expert system Li, Ilse Sau Chin


Sheet breaks are of major concern because they diminish the reliability and efficiency of the pulp production process, incur substantial cost, and pose a significant safety risk to plant operators. Consequently, this thesis addresses the problem of predicting and preventing sheet breaks by using Projection to Latent Structures (PLS) to model the sheet break process, and Principle Component Analysis (PCA) to monitor variables and define a region of healthy machine operation. An expert system is then employed to monitor major process variables, trouble shoot problems, and provide remedial action to prevent the occurrence of sheet breaks. Using data provided by the Weyerhaeuser Canada Pulp Mill in Grand Prairie, PLS modeling is able to reduce 43 measured process variables to three latent vectors that effectively reflect pulp quality and the operation of the headbox, presses and dryer. As such, the PLS model is able to predict 92% of sheet breaks. In addition, the PLS model is used to investigate and pinpoint key variables that are major contributors to sheet breaks, which in this pulp mill include: lumpbreaker loadings, dryer steam pressure, and dryer differential pressure. PCA analysis is then used to define a healthy region of machine operation, and subsequent tests using the model confirmed that when sheet breaks occurred, the scores indeed fell out of the healthy region. From the test set, PCA is able to predict 9 out of 12 sheet breaks. Additional analysis of the stock also reveals that fluctuating consistency of the stock in the headbox causes fluctuations in basis weight and consequently diminishes the pulp quality. The results from PLS analysis regarding variables that cause sheet break, and results obtained from PCA monitoring regarding fault detection are integrated into an expert system. The expert system effectively monitors the pulp production process, trouble shoots the situation when variables trend outside a healthy region of operation, and also suggests an appropriate course of action. Consequently, the expert system developed in this thesis functions to predict and prevent sheet breaks, and ultimately allows improved reliability, efficiency, quality, economics and safety of the pulp production process.

Item Media

Item Citations and Data


For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.