A mathematical model has been developed for the crystallization of aluminum alloy granules under cooling conditions in an aqueous and water-steam media. The practical significance of the mathematical model lies in predicting the average value of the dendritic parameter of the obtained granules depending on the granulation method, the characteristics of the granulation process and the size of the obtained granules. The average value of the dendritic parameter makes it possible to predict the fineness of the granule structure and, consequently, the mechanical properties of the granular material. The mathematical model makes it possible to determine the speed of a melt drop in an aqueous media and takes into account the presence of a steam jacket effect, i.e., a vapor layer that appears between a crystallizing drop and an aqueous media, which significantly reduces the intensity of heat removal and the rate of crystallization. The application of the mathematical model was tested on the obtaining of granules of high-alloyed aluminum alloys (alloys D1 and D16 of the Al–Cu–Mg system, alloys V95 and V96Ts of the Al–Zn–Mg–Cu system), made by centrifugal spraying of the melt from a perforated rotating crucible and drip method during cooling in an aqueous media. The cooling rate and the crystallization rate of the granules obtained in real experiments were determined by measuring the dendritic parameter of the material structure. The mathematical model showed a high convergence of the simulation results and of real experiments of aluminum alloy granulation.