The prediction of the dynamic behaviour of a chemical process is important for reactor design, optimization and safety. It is, however, beset with uncertainties of both, models and their input data. The latter are addressed here and the influence of uncertainties of key parameters, i.e. the heat of reaction, the reaction rate constant and the apparent energy of activation, on the calculation results and the conclusions drawn from them is shown. The conventional approach for the propagation of uncertainties through calculations, the Monte-Carlo method, is compared with calculations using polynomial chaos. The latter require considerably less time for calculation and are hence better suited for parameter variations, which are always needed in the design process. Both approaches are applied to an existing plant for manufacturing the explosive hexogen and illustrated by showing the evolution of the concentration of the product with time and the associated uncertainties. The ranges of predicted production quantities and raw material consumption as well as the impact of uncertainties on designing the dumping system for preventing a runaway reaction after cooling failure are also presented. (C) 2007 Published by Elsevier B.V.