The bounded-error approach to parameter estimation, mainly developed in the
context of control and signal processing, is applied in the electrochemist
ry field in order to obtain reliable estimates for kinetic parameters. The
method is based on the assumption that an uncertainty bar is available for
each measurement. A set guaranteed to contain all values of the parameter v
ector that lead to model outputs consistent with these error bars is then c
omputed, based on interval analysis and set inversion. The resulting techni
que is applied on simulated and experimental data for several classical ele
ctrochemical models. Its merits are compared to those of a more traditional
approach based on least square estimation by iterative local optimization.
A first obvious difference is that the point estimate provided by the latt
er method may or may not belong to the set estimated by the former, because
it does not take the bounds on the uncertainty into account. Moreover, as
our approach is global, it escapes the difficult problem of initialization
encountered with iterative optimization methods. Finally, the set obtained
is a natural characterization of the uncertainty on the estimated parameter
s. (C) 2000 Elsevier Science B.V. All rights reserved.