This article describes a numerical approach which allows for the analysis of the parametric identifiability of non-linear systems. It is shown that the analysis of global identifiability of N parameters of a system can be reduced to finding a global minimum of a specially designed 2N-dimensional function. We discuss the structure of this function and its physical interpretation. As an example we consider identifiability of a diffusion-reaction system and discuss experimental conditions for better identifiability of this system. (C) 2011 Elsevier Ltd. All rights reserved.