The objective of this work is to reduce batch-to-batch variations of k
appa number in batch pulping plants. We propose a model-based estimati
on approach that combines a nonlinear process model with on-line liquo
r measurements for estimation of key pulping states in the face of unk
nown feedstock variations. We show what measurements are needed and ho
w the estimation problem must be formulated in order to achieve suffic
iently fast recovery from the initial state/parameter errors. Simulati
on results indicate that, with the proposed estimator, target kappa nu
mbers can indeed be met very closely and significant reduction in the
batch-to-batch variability can be achieved.