Spray drying is an effective method to reduce shock sensitivity of energetic materials with the advantages of integrating atomization, drying, crystallization, and coating in one step. However, with the complex hydrodynamics and crystal growth process during spray drying, the development of theoretical models is very difficult. Therefore, five types of artificial neural network (ANN) models are proposed to accurately predict the mean particle size of energetic materials prepared by spray drying, such as Cascade-forward back propagation neural network (CFBP), Elman-forward back propagation neural network (EFBP), Feed-forward back propagation neural network (FFBP), Generalized regression neural network (GR) and Radial basis neural network (RB). The model input parameters are the inlet temperature (T), the liquid flow rate (L), the gas flow rate (G), the mass fraction (w), the molecular weight (M), and the surface tension (δ). The output parameter is the mean particle sizes (dp). To further illuminate the superior performance of ANN model, the effects of temperature, liquid flow rate, gas flow rate, mass fraction, and surface tension on the mean particle size are conducted. The ANN model of mean particle size for the energetic materials prepared by spray drying could be much useful for further improving its property.