The main aim of this work is to give a new approach of optimal control during the phase of identification. The question is how to tune the control action to be applied during the experiment optimize a criterion which is function of the sensitivities of the mesure with respect to the parameters of the model to be identified. This approach coupling constrained predictive controller and estimator solves on line the problem of identification by using the observer. In that sense, it is an approach allowing an optimal and automatic design of experiment, while performing at the same time the identification of one parameter of the specified model. The real time aspect was taken into account in the formulation of the solution. In this context, we introduced two strategies for optimal identification : the first one is based on a nonlinear model of prediction and the second one is based on a linear time varying model that may be used if the real time aspect becomes a critical parameter for the implementation of the approach. This approach of on line optimal identification was applied on two concrete problems in Chemical Engineering. These examples show the performance and the feasibility of this approach.
