Empirical and analytical approaches were used to model a pre-industrial pilot reactor to optimise a new decontamination process for AISI 304 L stainless steel. The alloy corrosion rate was modelled as an analytical function of the total cerium content and the ozone flow injected in the reactor. The empirical model of the corrosion rate, obtained by gravimetry, takes into account all the parameters, including the kinetics of alloy grains detached from the metal by intergranular corrosion. The discrepancy observed between the analytical and empirical models was explained by a two-step corrosion process. The dimensions and quantity of grains falling in the liquid were at first both underestimated. The study showed that the grains had to be extracted continuously from recirculating liquid in the future industrial reactor.