Prediction of the crystal´s growth rate based on BPNN and rough sets

Ammonium dihydrogen phosphate(DAP) is widely used in the industry. Usually this product contains impurities and substances and may not satisfy the need of modern agricultural and industrial demand, so we must dip and recrystalize for high quality product. In the process of crystallization, the nucleation rate and the growth rate are the most important parameter, then we study these parameter in order to instruct the technology of crystallization. A liquid fluidized bed crystallizer was used to determine the nucleation rate and the growth rate of ammonium dihydrogen phosphate (ADP) at crystallization temperature 15 and 25 for different saturation temperature solution. The growth rate in a Liquid fluidized bed is decided manly by the supersaturation, cooling temperature, saturation temperature and suspension density. In the paper we build a model predicting the growth rate through these conditions based on Back Propagation (BP) neural network with experimental data as training data. The experimental data, which collected from a Liquid fluidized bed, is preprocessed using the level of consistency in rough sets theory before be using as training sets in modeling process. The simulation results show that the neural network model given in this paper is capable of forecasting the behavior of growth rate exactly and rapidly, and the maximum relative error does not exceed 4.8% as compared with measured values. It also indicates the BP network has prodigious practicability.