Abstract
In the pulp and paper manufacturing process, numbers of material properties and end product qualities are changing steadily with time. Therefore it is considerably difficult to judge the variance of the production cost or the improved effect etc. through the operation data properly and fairly, because of their disturbances, when they are so tiny amount.
This paper describes a new method to evaluate the effect of the production cost and control effect etc. more accurately when a new control function was introduced or when process operation methods were changed for some purpose in the production processes.
Here, as an example, a case of a pulp bleaching process was taken up.
The proposed method is follows :
At first, find out the reliable multiple linear regression equation by using statistic analysis PLS (Partial Least Squares) regression method, for example, between the production cost as dependent variable and some primary explaining variables relates to the bleaching process. As it is well known, because the PLS analysis result is superior to the common regression analysis by MLR (Multiple Linear Regression) method in the stability and reliability. Sometimes, the regression coefficients calculated by MLR are tending to be inappropriate by the effects of over fitting or their collinearity.
Next, to get rid of the random variance and disturbance effects, correct and compensate the targeted “production cost” appropriately by “approximate linear correction method” so as to equalize the manufacturing conditions and incoming and outgoing material properties, utilizing the regression coefficients that were obtained by PLS regression model in advance.
As a result, it become possible to judge the control effect fairly and acceptably, even if the variance of the production cost is tiny change or the each explaining variable’s operating condition levels were changing intricately with each other.
