Parameters soft-sensing based on neural network in crystallizing process of cane sugar

Due to the online measurement difficulties of some parameters in the crystallizing process of cane sugar, parameter soft-sensing methods based on neural network are proposed. We provide a back-propagation neural network and a recurrent neural network to respectively build the density of boiling sugar juice and the speed of sucrose crystallizing soft-sensing models. The simulation results are allowed to carry out comparisons of running time, approximation capability and generalization capability between these two kinds of network. The results suggest that these two kinds of soft-sensing models based on neural networks are all able to provide good approximations to actual process. Finally, we discuss the influence of sample data on our soft-sensing models.