This paper deals with the experimental model based predictive control of the infrared cure cycle of a powder coating. It is based on a dynamic infinite dimensional model of the cure in one spatial domain, which aims to represent the evolution of the temperature and the degree of cure during the cure under infrared flow. The sensitivity of this model with respect to the main radiative property is experimentally highlighted under open loop conditions. This partial differential equation model is then approximated in finite dimension in order to be used by the predictive controller. Since the sampling time is small (one second), a special model predictive control formulation is used here, which aims to decrease the on-line computational time required by the control algorithm. Experimental evaluation of this controller that is based on the MPC@CB software is then presented. For black and white paintings, the robustness of this control algorithm is shown during an experimental temperature constrained trajectory tracking, even under a strong modeling uncertainty. The conclusion of this study is that this controller may be used for advanced control of powder coating cure. (c) 2009 Elsevier Ltd. All rights reserved.