Past efforts to use model-based controls in the paper industry have no
t been successful due to the sensitivity of the model control algorith
ms to sensor or model error and to noise. To better understand the rea
sons for this sensitivity and to develop more robust techniques, we ha
ve developed a series of dynamic models at various levels of control c
omplexity. A single-stirred tank model is used to illustrate the distu
rbance rejection and set point tracking of various control loops. A dy
namic two-ply liner paper machine model is used to show the systems-wi
de effects of individual control loops during upsets such as sheet bre
aks and the response to changes in basis weight. Feedforward and selec
tive control strategies proved to be essential for stable control of f
iber flow to the machine and, therefore, to controlling basis weight.
Improved numerical techniques were also needed for stable solutions of
the dynamic equations. The models have been integrated with a real-ti
me database, making impossible to run the system faster than real-time
with current input from the DCS. In this mode, the model could provid
e optimal control to portions of the paper machine to minimize the eff
ects of errors in consistency and the stability problems associated wi
th feedback. The results shown here merely illustrate the variety of c
ontrol concepts that could be evaluated.