Although jetmilling is a very energy consuming grinding process it is increasingly used in industry because very fine grinding product with a narrow size distribution is attained without contamination as the milling occurs by inter particle collisions. At Delft University a project has been started to achieve a considerable energy reduction of the milling process. Main feature of the improved grinding system will be an operation control based on in-line particle size measurement using the laser diffraction technique. The pilot plant is equipped with an external classifier for the removal and recirculation of larger particles leaving the jet mill. Phenomena occurring in the mill are not yet well modeled and thus no real optimization of the grinding process can be achieved. Operation conditions of jet mills are still determined by trial and error. This paper is to form the basis of a model which allows a dynamic modeling of the behavior of the jet mill/classifier system. The closed loop grinding system has been split in 3 subsystems (zones), two internal to the mill and one external through which the particles pass. Exchange of particles is not described by complicated residence time distribution functions but by ＇＇particle size＇＇ dependent probability functions.