In this study, methods for estimating kinetic parameters from on-line measu
rements of batch reactions are developed. The basic idea is to combine non-
linear parameter estimation with implicit calibration between measured spec
tra and concentrations given by a known kinetic model. Several alternatives
for the calibration step are discussed and the three most promising method
s have been studied more closely. The basic ideas and identifiability quest
ions are elucidated by simulation studies with simple kinetic models. The s
imulations show that the implicit calibration method works also with uncali
brated chromatograms. Two real esterification reactions with spectroscopic
data are studied. In the first one, concentration data are available and ou
r method gives results quite similar to those obtained in more traditional
ways. The latter reaction, however, was a fast one and consequently off-lin
e calibration based on known mixtures failed. In contrast, our method gave
almost identical parameter values to the ones obtained by an independent st
udy on the same reaction based on a complicated on-line chromatographic mea
surement system. In this case the results were compared with a new way of u
sing pure component spectra and generalized ridge regression with stoichiom
etric constraints. Our studies show that the most important factors in obta
ining reliable results are proper design of the batch experiments, proper w
eighting of the initial measurements of each batch and proper measure for g
oodness of fit. Copyright (C) 2001 John Wiley & Sons, Ltd.