A sequential design strategy for selecting experimental runs to obtain model discrimination and precise parameter estimation is tested via a simulation study of propylene oxidation kinetics. The strategy is used to design all runs including the preliminary ones which were arbitrarily chosen by earlier researchers. To design initial runs, crude initial parameter guesses may be used in the rival models until least squares estimates can be calculated. Even under conditions of very bad initial guesses and high error variances, this procedure selects whichever model is the correct one and estimates with precision its parameters, in fewer runs than previously reported.On a teste par simulation des cinetiques d'oxydation du propylene une strategie de calcul sequentiel pour la selection des essais experimentaux en vue d'obtenir une discrimination des modeles et une estimation precise des parametres. Cette strategie est utiliste pour concevoir tous les essais, y compris les premiers, choisis arbitrairement par d'autres chercheurs. Pour concevoir les essais initiaux, on peut faire appel a des estimations approximatives de parametres initiales dans les modeles concurrents jusqu'a ce que les estimations au sens des moindres carrees puissent itre calculdes. M&me en cas d'estimations initiales tres mauvaises et de variances d'erreurs importantes, cette methode permet de selectionner quel sera le bon modele et d'estimer avec precision ses parametres, et ce avec moins d'essais que mentionnes jusque la.Keywords: experimental design, initial runs, model discrimination, parameter estimation, propylene oxidation.here are two main objectives when using a statistical T design strategy in chemical kinetic modeling: (i) to establish a preferred rate mechanism for the reaction under study -this involves selecting several models based on different assumptions, and then performing experimental runs to discriminate among the various models, and (ii) to obtain the precise parameter estimates in the preferred model. To discriminate among several suggested models, carehlly chosen experimental conditions are needed. Arbitrarily collected data require too many runs and may even lead to the wrong model being chosen, as was found in the study by Froment and Mezaki (1970). First, a feasible experimental region of the several factors, from which experimental runs will be conducted, is selected. Then, statistical techniques can be used to chose the run conditions which lead to the preferred model most expeditiously.A Joint Design Criterion (JDC) developed by Hill et al. (1968) solves both problems of model discrimination and precise parameter estimation simultaneously in sequential m s . However as presented, the JDC designs runs only after a set of preliminary runs are conducted and analyzed. Sane et al. (1973), proposed a modified form of the JDC which designs all runs sequentially starting from the first one. With their Initial-runs JDC, crude guesses could be used as the initial parameters. Sane et al. (1973) demonstrated that desigmng the i...