A statistical model of aluminum agglomeration during combustion of solid composite rocket propellants is considered; the model describes the process dynamics, beginning from propellant heating in the combustion wave and ending by separation of agglomerates from the burning surface. An algorithm of computing the agglomeration process by the Monte Carlo method is proposed. A series of computations of aluminum agglomeration is performed; the density distribution functions for agglomerate sizes are derived; the dependence of the mean-mass size of agglomerates on the mean-mass size of ammonium-perchlorate particles is determined. The model proposed predicts power dependences of the mean-mass size of agglomerates on pressure and burning rate, which agrees with available experimental data.