Abstract A semi-physical auto-regressive moving average model of bending stiffness of the board produced at Assi Doman-Frovifors Bruk AB, as a function of measured control and disturbance variables, was identified. Based on the bending stiffness model, an adaptive on-line bending stiffness index predictor was implemented and found to have an RMS-error within the laboratory measurement accuracy. The predictor has been running for several months and consistently has a prediction accuracy of more than 75% within ± one standard deviation, for all product grades. The main difficulty in developing the predictor was that bending stiffness is seldom measured. The predictor will form the basis of a model predictive regulator of the bending stiffness.
[1]
Torsten Bohlin,et al.
Interactive System Identification: Prospects and Pitfalls
,
1991
.
[2]
Manfred Morari,et al.
Model predictive control: Theory and practice - A survey
,
1989,
Autom..
[3]
Lennart Ljung,et al.
System Identification: Theory for the User
,
1987
.
[4]
Alf Isaksson,et al.
Identification of ARX-models subject to missing data
,
1993,
IEEE Trans. Autom. Control..
[5]
Per Olof Gutman,et al.
A linear programming regulator applied to hydroelectric reservoir level control
,
1986,
Autom..